SCIENCE CHINA PRESS SCIENCE CHINA Life Sciences

[Link] [Link]
[Link]
REVIEW

Advances and applications in single-cell and spatial genomics

Jingjing Wang1†, Fang Ye1†, Haoxi Chai5†, Yujia Jiang9,10†, Teng Wang7,12†, Xia Ran1,4†, Qimin Xia7†, Ziye Xu6†, Yuting Fu1,2†,
Guodong Zhang1,2†, Hanyu Wu1,2†, Guoji Guo1,2,3,4#*, Hongshan Guo1,4*, Yijun Ruan5*, Yongcheng Wang6*, Dong Xing7,8*,
Xun Xu9,10,11* & Zemin Zhang7*
1Bone Marrow Transplantation Center of the First Affiliated Hospital & Liangzhu Laboratory, Zhejiang University School of Medicine, Hangzhou 310058, China
2Center for Stem Cell and Regenerative Medicine, Zhejiang University School of Medicine, Hangzhou 310058, China
3Zhejiang Provincial Key Lab for Tissue Engineering and Regenerative Medicine, Dr. Li Dak Sum & Yip Yio Chin Center for Stem Cell and Regenerative Medicine, Hangzhou 310058, China
4Institute of Hematology, Zhejiang University, Hangzhou 310000, China
5Life Sciences Institute and The Second Affiliated Hospital, Zhejiang University, Hangzhou 310058, China
6Department of Laboratory Medicine of The First Affiliated Hospital & Liangzhu Laboratory, Zhejiang University School of Medicine, Hangzhou 310058, China
7Biomedical Pioneering Innovation Center (BIOPIC) and School of Life Sciences, Peking University, Beijing 100871, China
8Beijing Advanced Innovation Center for Genomics (ICG), Peking University, Beijing 100871, China
9BGI Research, Shenzhen 518083, China
10BGI Research, Hangzhou 310030, China
11Guangdong Provincial Key Laboratory of Genome Read and Write, BGI Research, Shenzhen 518083, China
12Academy for Advanced Interdisciplinary Studies, Peking University, Beijing 100871, China

†Contributed equally to this work

#Lead contact author
*Co-corresponding authors listed in an alphabetical order of surnames (Guoji Guo, email: ggj@[Link]; Hongshan Guo, email: guohs@[Link]; Yijun Ruan, email: yjruan@[Link]; Yongcheng
Wang, email: yongcheng@[Link]; Dong Xing, email: dxing@[Link]; Xun Xu, email: xuxun@[Link]; Zemin Zhang, email: zemin@[Link])
Received 20 August 2024; Accepted 10 October 2024; Published online 20 December 2024

The applications of single-cell and spatial technologies in recent times have revolutionized the present understanding of cellular states and
the cellular heterogeneity inherent in complex biological systems. These advancements offer unprecedented resolution in the examination
of the functional genomics of individual cells and their spatial context within tissues. In this review, we have comprehensively discussed
the historical development and recent progress in the field of single-cell and spatial genomics. We have reviewed the breakthroughs in
single-cell multi-omics technologies, spatial genomics methods, and the computational strategies employed toward the analyses of single-
cell atlas data. Furthermore, we have highlighted the advances made in constructing cellular atlases and their clinical applications,
particularly in the context of disease. Finally, we have discussed the emerging trends, challenges, and opportunities in this rapidly evolving
field.

single-cell | technologies | cell atlas | multi-omics | spatial | clinical application

Introduction Chapter 1. Overview of single-cell and spatial

genomics
The human body is a highly organized system of approximately Conventionally, the genetic makeup of cells has been analyzed at
37 trillion cells encompassing hundreds of cell types (Han et al., the molecular or cellular level with limited markers. Conven-
2020; Jovic et al., 2022; Wen and Tang, 2022). Despite tional bulk-cell-based next-generation sequencing (NGS), also
originating from a single zygote, cells continue to accumulate known as bulk sequencing, analyzes cells from bulk tissues as a
genetic and epigenetic mutations as they divide and grow whole and provides averaged signals from heterogeneous cell
(Takeshima and Ushijima, 2019; Zong et al., 2012). In fact, cells populations. Through bulk sequencing, our understanding of
within the same tissue, organ, or cell type may contribute biology has revolutionized from relevant pathways to key genes,
differently to physiological or pathological processes. Under- but it is stuck at the molecular level as the differences between
standing this heterogeneity at the single-cell level is crucial for individual cells are not inquired. Low-throughput techniques
uncovering insights into developmental biology, disease mechan- such as single-cell polymerase chain reaction (PCR) and
isms, and therapeutic strategies. Analyzing the human body at immunofluorescence enable single-cell analyses, albeit only a
the level of individual cells can facilitate data integration and limited number of markers can be quantified this way (Kaur et
modeling across diverse tissues, thereby advancing our compre- al., 2019). Thus, there is an urgent need to develop single-cell
hension of human systems. sequencing technologies that can systematically profile indivi-

Citation: Wang, J., Ye, F., Chai, H., Jiang, Y., Wang, T., Ran, X., Xia, Q., Xu, Z., Fu, Y., Zhang, G., et al. (2025). Advances and applications in single-cell and spatial genomics.
Sci China Life Sci 68, 1226–1282. [Link]

SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 © Science China Press 2024
dual cells. Chapter 2. Single-cell sequencing technologies
To sequence single cells, it is a prerequisite to isolating single
So far, single-cell sequencing technologies have evolved from
cells from heterogeneous cell populations (Walker and Parkhill,
transcriptome to genome, epigenome, and proteome, from
2008). Major challenges encountered in single-cell isolation are
singular modality to multi-omics, from single-cell to subcellular,
the scalability and sensitivity of the isolation methods while
and scaled up from dozens of cells to millions of cells (Figure 1).
ensuring the quality and purity of the individual cells (Gross et
al., 2015). To tackle these challenges, various approaches have
been adopted, which are mainly based on the following three Single-cell RNA-seq
principles (Hu et al., 2016a; Kaur et al., 2019): (i) physical scRNA-seq estimates the RNA molecules of individual cells and
features, such as the size, density, electric charges, and provides a snapshot of the entire transcription landscape within
deformability; (ii) biological properties, such as the type of the assayed cell to reveal the genes that are being actively
surface proteins; (iii) molecular barcoding by using a split-pool transcribed into individual cells. Based on this information, the
ligation-based strategy. Considering the various aspects of cell cell identity, state, and type can be defined at a more granular
input/output, automation level, scalability, and cell viability after level (Cuomo et al., 2023). In a diploid cell, many expressed genes
isolation, four major approaches have been adopted and widely have more than a dozen copies of the mRNAs, which makes it
applied across laboratories, including micromanipulation, laser- relatively feasible to quantify them in individual cells (Wen and
capture microdissection (LCM), fluorescence-activated cell sort- Tang, 2022). For this reason, perhaps, scRNA-seq stands out as
ing (FACS), and microfluidics-assisted droplet encapsulation. Of the earliest, most commonly-used, and most cutting-edge
these approaches, microfluidics-assisted droplet encapsulation (especially in terms of throughput (Tanay and Regev, 2017)
exhibits better throughput performance and automation (Zhou et technology that has been developed for single-cell sequencing.
al., 2021). The inaugural demonstration of scRNA-seq on a NGS platform
The second prerequisite of single-cell sequencing is to dates back to 2009 (Tang et al., 2009). This first model analyzed
effectively amplify trace amounts of genetic materials from single the whole transcriptomic characterization of manually isolated
cells (Walker and Parkhill, 2008). PCR with random or semi- single cells. Since then, scRNA-seq has scaled up rapidly
random primers has laid the groundwork for single-cell (Svensson et al., 2020)—from only 8 cells in the first study to
amplification. Although it is bias-prone, PCR is widely used and more than 11 million cells in a recent publication (Qiu et al.,
later optimized across diverse hybridization-based or isothermal 2024). Over the ensuing nearly 15 years, the landscape of
amplification methods (Kaur et al., 2019). The hybridization- scRNA-seq sequencing technologies has witnessed an unprece-
based methods encompass methods that have been modified from dented surge in terms of development and diversification (Table
the conventional PCR, including methods such as linker-adaptor S1). The conventional scRNA-seq workflow encompasses crucial
PCR, primer extension preamplification PCR, and degenerate stages, including single-cell separation, library construction,
oligonucleotide-primed PCR. To overcome the limitations of high-throughput sequencing, and subsequent data analysis. The
hybridization methods (e.g., low coverage, allele dropouts, and common steps involved in transcriptome library generation
amplification bias), the isothermal methods and in vitro include reverse transcription (RT) into first-strand cDNA, second-
transcription (IVT) were subsequently developed (Kaur et al., strand synthesis, and cDNA amplification. Notably, diverse
2019; Walker and Parkhill, 2008). scRNA-seq technologies exhibit distinct advantages and limita-
Developed in parallel with single-cell genomic technologies, tions throughout the process. Although certain preliminary
spatial genomic methods preserve the position information and technologies have now turned obsolete, some others have
map the spatial organization of gene expression and other undergone continuous refinements, such as the markedly
molecular features within tissues. Exciting progress has been enhancing pace, throughput, sensitivity, coverage, capacity for
recorded for spatial transcriptomics (STs), which combines long reads, and overall consistency. Pioneering technologies
spatial information with transcriptomics, thereby allowing gene such as Smart-seq (Ramsköld et al., 2012), introduced in 2012,
expression profiling while preserving tissue architecture. In have now evolved into more advanced iterations such as the
addition, multiplexed in situ hybridization detects multiple RNA recently updated Smart-seq3xpress (Hagemann-Jensen et al.,
or DNA targets within a single tissue section and provides high- 2022). In plate-based scRNA-seq techniques such as Smart-seq
resolution spatial maps of gene expression. and CEL-Seq, the experimental cells are physically separated into
With technological advances in single-cell barcoding and micro-well plates (such as a 96-well or 384-well plate) in a one-
amplification, single-cell and spatial genomics has emerged, cell one-well fashion, with only a few dozen or hundred cells
wherein cells can be profiled with both molecular and position analyzed at a time. However, in the droplet-based strategy, such
indexes in a high-throughput manner. Recently, these technical as Drop-seq and inDrop, single cells are encapsulated into
forces have contributed to the construction of comprehensive microdroplets containing beads that are prebarcoded by unique
cellular atlases involving mapping of the molecular character- oligonucleotides. After cell lysis, mRNA molecules are hybridized
istics of all cell types within an organism. Further mapping of and reverse transcribed with the barcodes and then eventually
cellular changes in various diseases can provide unprecedented sequenced using labels that correspond to their cell origin
insights into disease mechanisms and potential therapeutic (Cuomo et al., 2023). Notably, only the 5′ or 3′ end of transcripts
targets. However, to fully realize the potential of these atlases, is sequenced and thousands of cells are analyzed simultaneously.
we need to overcome several hurdles. Here, we have discussed The pivotal moment in scRNA-seq’s commercialization tran-
some of the key areas in the field of single-cell and spatial spired with the launch of 10x Genomics’ Chromium in 2015,
genomics identified in the past decade and with potential in the which provides a benchtop solution accessible to scientists and
future. catalyzing advancements in other commercial scRNA-seq

[Link] SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 1227
Figure 1. Overview of single-cell technologies in different omics.

technologies (Zheng et al., 2017b). Together, these events have precise dissection of single cells from tissues with spatial accuracy
helped to shape the dynamics and rapidly evolving landscape of (Chen et al., 2017b; Nichterwitz et al., 2016). A microfluidic
single-cell transcriptomics. device based on micropillar arrays offers highly efficient isolation,
detection, and collection of low-abundance cells (e.g., circulating
Single-cell separation tumor cells (CTCs)) (Chen and Wang, 2023). Minimally invasive
As the primary distinction from conventional sequencing nanotweezers allow spatially controlled extraction of samples
methods, the separation of single cells from samples is a crucial from living cells with single-molecule/organelle precision (Na-
step in gathering transcriptome information from individual dappuram et al., 2019).
cells. Conventional techniques for single-cell separation have An additional avenue within the design of single-cell separa-
been thoroughly scrutinized in other studies (Hwang et al., 2018; tion methods seeks to optimize both throughput and efficiency.
Pensold and Zimmer-Bensch, 2020; Tan et al., 2010), including The release of the first single-cell automated prep system,
limiting dilution, micromanipulation, LCM, and FACS. In low- Fluidigm C1, which can be implemented with CEL-Seq2
throughput scRNA-seq technologies (e.g., Smart-seq2 (Picelli et (Hashimshony et al., 2016) and Smart-seq/C1 (Shalek et al.,
al., 2013), single-cell universal poly(A)-independent RNA 2014), signified a significant advancement. Microwell-based
sequencing (SUPeR-seq) (Fan et al., 2015b), molecular crowding scRNA-seq methods (e.g., CytoSeq (Fan et al., 2015a), Seq-Well
single-cell RNA barcoding and sequencing (mcSCRB-seq) (Bag- (Gierahn et al., 2017), and Microwell-Seq (Han et al., 2018b)),
noli et al., 2018), and massively parallel single-cell RNA- wherein cells are deposited into wells by gravity at random, offer
sequencing (MARS-Seq) (Jaitin et al., 2014)), the single cells the advantages of low cost and high throughput. The introduc-
obtained via these techniques can be individually segregated into tion of droplet microfluidic technology has revolutionized cell
tubes, which subsequently undergo distinct reactions. Certain throughput (Wen et al., 2016). Basically, it enables the
conventional low-throughput methods for single-cell separation encapsulation of single cells and reagents in independent
have now evolved to facilitate detailed examination of intricate aqueous microdroplets (typically tens to hundreds of μm in
solid tissues, rare samples, and even subcellular fractions. For diameter) in a high-throughput manner (e.g., Drop-seq (Macosko
instance, LCM utilizes a computer-aided laser system for the et al., 2015), inDrop (Klein et al., 2015) and 10x Genomics’

1228 SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 [Link]
Chromium). Nevertheless, both microwell- and droplet-based The cellular origin of RNA is labeled with specific-barcoded RT
scRNA-seq methods rely on stochastic pairing of cells and primers, e.g., MARS-Seq, Drop-seq, CytoSeq, and vast transcrip-
barcoded beads, which is constrained by the mathematical tome analysis of single cells by dA-tailing (VASA-seq) (Salmen et
principle of Poisson statistics, resulting in low utilization of both al., 2022). To further facilitate scalable profiling of single cells,
the cells and beads. To resolve these limitations, innovative cost-effective and high-throughput scRNA-seq methods (e.g., sci-
approaches such as dTNT-seq (Bai et al., 2020), which integrates RNA-seq, SPLiT-seq, and sci-RNA-seq3 (Cao et al., 2019)) have
dielectrophoresis (DEP)-trapping-nano well-transfer, and Well- been developed, wherein cells are combinatorial indexed by RT,
paired-seq (Yin et al., 2022), which consists of thousands of size cDNA ligation, and PCR. The majority of droplet- and microwell-
exclusion and quasi-static hydrodynamic dual wells, have been based scRNA-seq technologies incorporate microparticles or
developed to enhance microreactor and cell/bead capture hydrogel microspheres (beads), each with distinct barcoded
efficiency. Single-cell combinatorial indexing (sci), incorporating primers. Each barcoded bead is coencapsulated with an
in situ nucleic acids indexing and a “split-pool” strategy for individual cell and reaction reagents utilizing microfluidic chips.
unique labeling, serves as a low-cost methodological framework However, the diversity of such barcoded beads presents a
for high-throughput scRNA-seq (e.g., sci-RNA-seq (Cao et al., considerable challenge for these bead-based high-throughput
2017) and SPLiT-seq (Rosenberg et al., 2018)). Datlinger et al. scRNA-seq methods. Drop-seq introduced a 12-round “split-and-
(2021) introduced a “single-cell combinatorial fluidic indexing” pool” strategy to directly synthesize oligonucleotide primers on
(scifi) strategy and developed scifi-RNA-seq—a method that beads. inDrop and Microwell-Seq synthesized the library of
combines one-step combinatorial preindexing of entire transcrip- barcoded hydrogel microspheres using two or three split-pool
tomes with subsequent single-cell RNA-seq using microfluidics— rounds of hybridization and primer extension. Generally, among
which eliminated the need for absolute single-cell separation and all cell barcoding methods, the microwell- and nanowell-based
overcame the limitations of Poisson distribution in encapsula- methods demonstrate comparable sensitivity, throughput, and
tion. These limitations of single-cell separation methods originate lower cost when compared with the droplet platform method.
from the essential use of microwells or microfluidic devices The stability of commercialized equipment in both droplet and
necessary for sample processing. The microfluidics-free particle- microwell platforms has provided diverse choices for users based
templated instant partition sequencing (PIP-seq) (Clark et al., on the sample size, cell size, and budget. The droplet platform
2023) allows single-cell encapsulation and barcoding of cDNA in may not be appropriate for large cells that can lead plugs. On the
uniform droplet emulsions by using only a vortex. Recently, an other hand, pool-split indexing methods are convenient and
instrument-free scRNA-seq method RevGel-seq (Komatsu et al., require no special equipment, making them suitable for almost all
2023), which utilizes a reversible hydrogel, has presented a novel laboratories (does not need an automatic pipetting system).
cell-bead pairing strategy that does not require nanowells or Nonetheless, there is a drawback to the need for manual pipetting
droplets, thereby introducing a new direction in the development proficiency to ensure reproducible outcomes.
of single-cell separation in scRNA-seq technologies.
Droplet microfluidic technology also offers the benefits of cDNA library generation strategies
adaptability and transportability in high-throughput scRNA-seq. Together with the enhancements in throughput, the cDNA
Various cell characterization technologies can be incorporated library generation of scRNA-seq technologies has incessantly
with microfluidic chips to further select specific cells during evolved to amplify the sensitivity and accuracy of RNA in single
single-cell separation, which can be achieved by regulating the cells. This aspect remains the most salient distinction among
micron-diameter channels’ liquid flow. Computational Sorting various scRNA-seq methods. In the following section, we have
and Mapping of Single Cells (COSMOS) is a platform based on reviewed these diverse strategies.
artificial intelligence (AI) and microfluidics that can characterize Before the library generation reactions, some of these scRNA-
and sort single cells based on high-dimensional embedding seq technologies (e.g., droplet-based Drop-seq and microwell-
vectors of morphology without needing biomarker labels and based CytoSeq) lyse single fresh cells in separated chambers and
stains/dyes (Salek et al., 2023). In addition, an automated capture the released RNA molecules through hybridization using
Raman-based sorting method involves the use of optical tweezers primers. Some scRNA-seq technologies (e.g., plate-based SPLiT-
and microfluidics to sort individual cells of interest (Lee et al., seq, microwell-based Microwell-seq 2.0 (Chen et al., 2021a),
2021). Digital microfluidics (DMF) is also an emerging technique Microwell-seq3 (Ye et al., 2024b), and droplet-based snRandom-
that employs dielectric wetting to precisely manipulate discrete seq (Xu et al., 2023f)) are compatible with fixed cells and nuclei,
fluids, allowing efficient and non-destructive isolation of single allowing the capturing of RNA molecules and the generation of
cells (Zeng et al., 2024). cDNA through an in situ RT reaction.
scRNA-seq methodologies can be categorized into three major
Cell barcoding types—oligo-dT, random/not-so-random primer (NSRs), and
A significant advancement that substantially elevates the probe-based—depending on the primers employed for the initial
enumeration of profiled single cells involves the introduction of cDNA synthesis. Ever since the first scRNA-seq of Tang et al.
single-cell tagged reverse transcription (STRT-seq) (Islam et al., (2009) a significant proportion of classical scRNA-seq technol-
2011), which uses cell-specific barcodes allowed for multiplexed ogies use oligo-dT primers for the first-strand cDNA synthesis.
single-cell RNA-seq on the Illumina platform. Barring these low- Oligo-dT priming is an efficient approach to capturing mRNAs
throughput scRNA-seq technologies (e.g., Smart-seq2, Smart- with minimal uninformative reads (e.g., those derived from
seq/C1, SUPeR-seq, multiple annealing, and dC-tailing-based rRNAs). The state-of-the-art oligo-dT-based scRNA-seq methods
quantitative single-cell RNA-seq (MATQ-seq) (Sheng et al., (e.g., plate-based Smart-seq2, droplet-based 10x Genomics’
2017)), virtually all scRNA-seq technologies utilize cell-specific Chromium, and microwell-based CytoSeq) are adequately
barcodes, each employing distinct application methodologies. sensitive to quantify and determine cell states with high

[Link] SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 1229
accuracy. Nevertheless, this strategy solely captures and primer with a barcode and an upstream primer-binding
amplifies polyadenylated RNAs to consequently yield an in- sequence, and the barcode is introduced into the 5′-end of the
complete depiction of the transcriptome. Moreover, the oligo-dT- transcripts by Moloney Murine Leukemia Virus reverse tran-
primed cDNA synthesis requires high-quality intact RNAs, scriptase (MMLV RT) (Zhu et al., 2001). Smart-seq epitomizes a
thereby precluding the application of oligo-dT-based scRNA-seq robust and precise methodology that enhances read coverage
technologies with samples that are deteriorated or degraded (e.g., across transcripts, with marked sensitivity and accuracy in
formalin-fixed paraffin-embedded (FFPE) samples). Random quantification. It also leans heavily on the terminal transferase
priming is another approach for the first-strand cDNA synthesis, and template-switching activities intrinsic to MMLV RT. It has
which can cover full-length of any size of transcripts. SUPeR-seq been hypothesized that template switching transpires sequen-
introduced random primers to enable the simultaneous detection tially, wherein once the RT reaction reaches the 5′-end of an
of both poly(A)+ and poly(A)− RNA species (e.g., circular RNAs). RNA molecule, there is an untempled appendage of three
MATQ-seq is more sensitive with the use of RT primers that protruding nucleotides (+CCC) to the 3′-end of the nascent
consist of the MALBAC primer, three consecutive G or T, and 20 cDNA, which then hybridizes with the complementary rGrGrG
consecutive T (dT20). RamDA-seq combined a novel RT 3′-end of the template-switching oligo (TSO), thereby facilitating
technology: RT with random displacement amplification (RT- RT to swap its template and resume transcription. The SMART
RamDA) and NSRs (Hayashi et al., 2018). The high-throughput strategy, which significantly enriches transcripts with intact 5′-
MATQ-drop (Niu et al., 2023), snRandom-seq (Xu et al., 2023f), ends and eliminates the second-strand synthesis, is the most
and snHH-seq (Chen et al., 2024a) also used random primers for feasible strategy to integrate with plate-, microwell-, or droplet-
total RNA capture and average gene body coverage. 10x based full-length scRNA-seq methods. Thus, the switching
Genomics’ Chromium Single Cell Gene Expression Flex applies mechanism at the 5′-end of the RNA template (SMART) has
probe-based chemistry for RNA capture in relatively low-quality become the most common strategy used by Smart-seq2, CytoSeq,
fixed samples. In the FFPE samples with worse RNA quality, the Drop-seq, and 10x Genomics’ Chromium. An IVT strategy based
random primer-based snRandom-seq displayed much higher on T7 RNA polymerase was developed for the synthesis of the
sensitivity when compared with other probe-based (snPATHO- second strand without any need for an anchor sequence. The IVT
Seq (Vallejo et al., 2022)) or poly(A)-based (snFFPE-seq (Chung strategy has been adopted by several scRNA-seq technologies,
et al., 2022)) scRNA-seq methods of FFPE samples. such as CEL-Seq (Hashimshony et al., 2012), CEL-Seq2, MARS-
Once the RNA molecule undergoes RT into cDNA, the seq, and inDrop. CEL-seq is the first method to employ IVT in
comprehensive amplification of the cDNA library is mandated scRNA-seq, which overcomes the limitation of the small initial
before its in-depth sequencing. Presently, scRNA-seq primarily amounts of RNA. In CEL-seq, the amplified RNA via IVT strategy
employs three strategies for cDNA library amplification, namely is fragmented to a size distribution appropriate for sequencing
PCR-based amplification, linear isothermal amplification via T7- and then converted into a library compatible with Illumina
based IVT, and amplification utilizing Phi29 DNA polymerase. As sequencing through ligation with the 3′ adaptor. However, the
the Phi29 DNA polymerase-based amplification has not been ligation step is not efficient and introduces primer dimers that
updated recently, we have reviewed scRNA technologies using interfere with sequencing. CEL-Seq2, a refined iteration of CEL-
PCR- and IVT-based amplification. Seq executed on Fluidigm’s C1 system, dispenses with the
Based on the synthesis pattern of second-strand cDNA with a ligation phase by incorporating the Illumina adaptor directly
second anchor sequence for PCR amplification, PCR-based during the second RT step, serving as a 5′-tail affixed to a random
amplification can be categorized into homopolymer tailing, hexamer. MARS-Seq provided an optional addition of plate
template-switching, and random priming. In the first scRNA- barcode by ligating the 3′ adaptor on the amplified and
seq research by Tang et al. (2009), ~30-nt polyA tails are fragmented RNA. In inDrop, the amplification of RNA according
attached to the first-strand cDNA at the 3′-end through terminal to CEL-Seq was conducted in nanoliter droplets. The recently
deoxynucleotidyl transferase. Subsequently, a second poly(dT) reported high-throughput VASA-seq also embraced this IVT
primer, anchored differently, is employed to procure double- tactic. However, it involves the engagement of a picoinjection
stranded DNA, which is subsequently amplified via PCR. apparatus for two successive operations in the droplets of VASA-
Subsequent Quartz-Seq (Sasagawa et al., 2013), MATQ-seq, seq, which poses a challenge to its broad application due to the
snRandom-seq, and snHH-seq also adopted a similar homo- increased complexity. Although Quartz-Seq uses a homopolymer
polymer tailing strategy. Considering that the polyA or polyC tail tailing strategy for the cDNA library amplification, IVT has been
can be added to any excrescent RT primers, the cDNA library will used in the Quartz-Chip method to synthesize labeled cRNA from
contain byproducts from the surviving primers. In Quartz-Seq, all the amplified cDNA for the microarray analysis.
the steps of the whole-transcript amplification are executed in a
single PCR tube. The preponderance of superfluous RT primer is Transcript coverage in cDNA library generation
digested by exonuclease I, while the amplification of byproducts Predicated on the transcript coverage, these scRNA-seq technol-
is inhibited through the application of suppression PCR primer. ogies present a clear demarcation into 3′-end, 5′-end, and full-
Conversely, in MATQ-seq, the excrescent RT primers are digested length modalities. Traditional NGS is restricted to abbreviated
by adding T4 DNA polymerase into the original reaction tube to transcript fragments, introducing a noticeable 3′-end bias in read
avoid the production of by-products. The droplet-based MATQ- distribution for high-throughput scRNA methodologies that use
drop, snRandom-seq, and snHH-seq are performed on fixed single oligo-dT priming, such as Drop-seq, inDrop, and 10x Genomics’
cells or nuclei, which allow indirect washing with buffer after the Chromium (Zhang et al., 2019d). Conversely, 5′-tag counting
RT reactions. STRT-seq was the first to introduce a template- scRNA-seq methods like STRT-seq and STRT-seq-2i (Hochgerner
switching strategy for second-strand cDNA synthesis. In STRT- et al., 2017) exhibited a bias toward the 5′-end of transcripts.
seq, mRNA is reverse-transcribed into cDNA by a tailed oligo-dT Although both 5′-end and 3′-end scRNA-seq demonstrated

1230 SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 [Link]
comparable cell typing capabilities, 5′-end approaches offer RNA dynamics
specific advantages in certain applications. For instance, C1 scRNA-seq of gene expression dynamics can provide unique
Cap analysis gene expression (C1 CAGE) (Kouno et al., 2019) insights into transcription perturbation during development,
employed an original sample multiplexing strategy in the C1 drug treatment, and disease progression. Time-resolved dimen-
microfluidic system for the detection of transcription start sites sion of RNA synthesis in the system can achieve temporal
and enhancer activity at single-cell resolution. In addition, resolution, which is important in an organism. In 2019, single-
scBCR-seq (Goldstein et al., 2019), which combines 5′-end cell, thiol-(SH)-linked alkylation of RNA for metabolic-labeling
scRNA-seq with single-cell V(D)J sequencing, profiled B cell sequencing (scSLAM-seq) (Erhard et al., 2019) and new
receptor repertoires. The ultra-high throughput 5′-end scRNA- transcriptome alkylation-dependent single-cell RNA sequencing
Seq method FIve PRime End Single-cell Combinatorial Indexing (NASC-seq) (Hendriks et al., 2019) utilizes nucleoside analog 4-
RNA sequencing (FIPRESCI) (Li et al., 2023d) combines in situ thiouridine (4sU) and iodoacetamide-based chemical conversion
preindexing using Tn5 transposomes with droplet template reactions to label new RNA during transcription. The abundant
switch oligo (TSO) barcoding, thereby allowing sample multi- T-C conversions in new RNA distinguish them from unlabeled
plexing and multimodal profiling of the entire transcriptome and “old” RNA. In 2020, sci-fate (Cao et al., 2020b) and scNT-seq
immune receptor repertoire. However, the 3′- or 5′-end bias (Qiu et al., 2020) reported similar metabolic labeling methods by
limits the mutation and splicing analysis (Chen et al., 2024a). using split-pool indexing and droplet microfluidics platform to
Smart-seq2 has been the gold standard for full-length scRNA-seq achieve high-throughput sequencing of RNA dynamics in single
owing to its sensitivity and robustness. Each specimen procured cells, respectively. In addition to these metabolic labeling
via the low-throughput oligo-dT-based Smart-seq2 undergoes methods, Liver-seq allows direct extraction of RNA in single cells
individual processing. The cDNA libraries of Smart-seq2, once through fluidic force microscopy and preserves cell viability
fragmented, are primed for thorough sequencing and deeper (Chen et al., 2022b). However, the equipment and throughput
analysis. This step unlocks opportunities for characterizing splice restrict its further application. Recently, single-cell nascent RNA
isoforms, allelic variants, and single-nucleotide polymorphisms sequencing (scGRO-seq) employed click chemistry rather than
at a single-cell resolution. High-throughput VASA-seq employs metabolic labeling to directly evaluate new RNA and to
oligo-dT primers as well but staggers and polyadenylates every transcribe RNA polymerases in single cells (Mahat et al.,
RNA molecule in a lysed single cell, thereby facilitating the 2024). The dynamic landscape of global transcription unveiled
apprehension of total RNAs via oligo-dT primers. Even SUPeR- co-transcription between pluripotency genes and their enhan-
seq, which uses random (AnchorX-T15N6) primers, shows mild cers.
3′-end bias. In fact, some scRNA-seq technologies like MATQ-seq,
snRandom-seq, and snHH-seq also incorporate random primers,
Single-microbe RNA-seq
thereby integrating a multiple annealing tactic during the RT
process for the capturing of full-length transcripts. Microbial traits exhibit both heterogeneity within populations
and precise transcriptional control, which emphasizes the need
Sample multiplexing in cDNA library generation for conducting single-microbe transcriptional analyses. However,
Pre-indexing based on separated tagmentation or RT provides an the existing scRNA-seq technologies face substantial incompat-
effective way for sample multiplexing. Multiplexing for distinct ibility with single-microbe RNA-seq due to several formidable
samples in one experiment can be achieved through antibody technical challenges. For instance, a prime constraint arises from
labeling, genetic barcoding, and click chemistry reactions. Cell the lack of 3′-end poly(A) tails in bacterial mRNAs, which are
Hashing (Stoeckius et al., 2018) used oligo-tagged antibodies significant capture regions for poly(T) primers. Moreover, the
against ubiquitously expressed surface proteins to mark cells RNA content in an ordinary bacterium is nearly two orders of
obtained from different samples. The mimic poly-A tails in the magnitude less compared with that in a standard mammalian
oligos were captured and analyzed in the sequencing data to cell (Westermann and Vogel, 2021), with ribosomal RNA (rRNA)
distinguish these samples. This method was then further constituting over 80% of total bacterial RNAs (Giannoukos et al.,
developed to simultaneously analyze target proteins and RNA 2012). The rigid cell wall of bacteria presents additional
at a single-cell level. In contrast, CellTag Indexing (Guo et al., challenges toward cell lysis and the subsequent extraction of
2019) used heritable predefined genetic barcodes to label RNA in droplets. Nevertheless, the emergence of recent technical
different cell types sourced from different samples, which allows enhancements has expedited the realization of bacterial single-
long-term cell tracking both in vitro and in vivo. The genetic cell transcriptomics. The primary procedure for single-microbe
barcoding strategies were adapted to gene expression perturba- RNA-seq significantly parallels the protocols devised previously
tion and lineage tracing. For sample multiplexing based on click for eukaryotic cells. For single microbe separation, in the early
chemistry reaction, the “ClickTags” method (Gehring et al., single-bacterium RNA-seq method, BaSiC RNA-seq (Wang et al.,
2020) involved rapid cross-linking reaction along with the 2015), individual bacterial cells underwent isolation, with
identification of methyltetrazine-modified DNA oligonucleotides subsequent purification and amplification of total RNA from
by using cross-linker NHS-trans-cyclooctene (NHS-TCO). An- each cell separately. Some relatively low-throughput methods for
other method, MULTI-seq, used lipid- and cholesterol-modified single-bacterium RNA-seq (such as Imdahl et al.’s method
oligonucleotides as an anchor with plasma membranes for cell (Imdahl et al., 2020) and Homberger’s method (Homberger et
labeling (McGinnis et al., 2019). In general, click chemistry al., 2023)) have been devised by employing FACS. Two high-
reactions can provide flexible and effective labeling approaches throughput bacterial scRNA-seq methods (prokaryotic expres-
for cells, nuclei, and molecules. Thus, the modifications of DNA sion profiling by tagging RNA in situ and sequencing (PETRI-seq)
and proteins can thereby be distinguished and sequenced by (Blattman et al., 2020) and microSPLiT (Kuchina et al., 2021))
other multi-omics methods. adopted a split-pool barcoding strategy, which was initially

[Link] SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 1231
established for eukaryotes to allow researchers to simultaneously amplified using isothermal PCR with random or semi-random
analyze thousands of bacteria. Recently, the commercially primers, such as multiple displacement amplification (MDA) and
available 10x Genomics droplet platform—a popular choice for multiple annealing and looping-based amplification cycle (MAL-
high-throughput single-cell RNA-seq—was embraced by probe- BAC). MDA provides a better but non-uniform genome coverage,
based bacterial sequencing (ProBac-seq) (McNulty et al., 2023), making it suitable for the detection of single-nucleotide variations
M3-Seq (Wang et al., 2023a), and BacDrop (Ma et al., 2023b). (SNVs). In comparison, MALBAC provides a reduced but more
M3-Seq combined plate-based in situ indexing by RT with droplet- even coverage and is thus more suitable for detecting copy
based indexing by ligation. Moreover, smRandom-seq (Xu et al., number variations (CNVs). To further minimize priming and
2023e) employed an enhanced barcoding platform, which was amplification biases, linear amplification via transposon insertion
originally devised for single-cell barcoding. This platform is (LIANTI) was later developed, which demonstrated the highest
characterized by smaller barcoded beads and droplets, as genome amplification uniformity (Chen et al., 2017a). In
complemented by a refined primer release and barcode synthesis addition to the conventional scDNA-seq method, some novel
method. For cDNA library generation, Imdahl et al. (2020) and technologies have been developed based on scDNA-seq. For
Homberger et al. (2023) developed methods based on the poly example, Leighton et al. (2023) developed a multipatient-
(A)-independent MATQ-seq strategy, which was previously targeted scDNA-seq that enables the construction of mutational
established for eukaryotes. In Homberger’s refined bacterial lineage in breast cancer.
single-cell RNA-seq method through automated MATQ-seq, the
enhancements in gene coverage and gene detection thresholds
Single-cell epigenome sequencing
improved, and small RNAs were successfully identified at the
single-cell level. PETRI-seq tagged RNA molecules in situ within a The epigenome comprises a set of heritable modifications to DNA,
fixed single microbe using random primers. microSPLiT poly- RNA, or histones within a cell, which can alter gene expression.
adenylated the mRNA in-cell with E. coli poly(A) polymerase I Considering that epigenetic information is generally dispersed
(PAP). The subsequent protocols for these two technologies were throughout the genome, single-cell epigenome sequencing poses
both built on SPLiT-seq. ProBac-seq captured individual tran- particular challenges (Wen and Tang, 2022).
scripts with pre-designed DNA probes complementary to protein-
coding sequences, which potentially limited unbiased discovery. DNA methylation
Recently, three high-throughput single-microbe RNA-seq meth- In mammals, chemical modifications to the DNA mainly involve
ods (BacDrop, smRandom-seq, and M3-seq) were developed the addition or removal of methyl groups from the C-5 position of
using random primers for in situ cDNA generation. The previous cytosine (5mC) or the N6 position of adenine (m6A). As the
single-microbe RNA-seq technologies, such as PETRI-seq, ex- dominant type of DNA modification, 5mC occurs mainly in the
hibited an abundance of signals from rRNA, which could form of 5′-3′ CpG dinucleotides (Liu et al., 2023b). Remarkably,
compromise mRNA detection and result in high sequencing DNA methylation has been demonstrated to play a critical role in
costs. For the mRNA enrichment process, microSPLiT used the development of human diseases; analysis of DNA methylation
mRNA-preferred PAP to polyadenylate RNA. BacDrop performed at the single-cell level is thus essential to reveal cellular
rRNA depletion in situ through probe-hybridization and followed heterogeneity in DNA methylation. The optimization of bulk
RNase H digestion. These in situ mRNA enrichment methods may DNA methylation sequencing technologies has accelerated the
pose a risk of losing unamplified non-rRNA transcripts. M3-seq development of a variety of single-cell DNA methylation
also adopted this probe-hybridization-based strategy but per- sequencing (scDNAm-seq) technologies, mainly based on restric-
formed rRNA depletion after library amplification. Another rRNA tion digestion or post-bisulfite adapter tagging (PBAT) (Liu et al.,
depletion strategy (CRISPR-based DASH (Gu et al., 2016)) was 2023b). The first scDNAm-seq technology was developed in
employed by Homberger’s method and smRandom-seq at the 2013 based on reduced representation bisulfite sequencing
sequencing library level. Although these singular microbe RNA- (scRRBS), which utilizes the strategy of one-tube enzymatic
seq methodologies have proven effective with both gram- reaction to reduce unnecessary DNA loss (Guo et al., 2013).
negative and gram-positive laboratory bacterial strains, the Subsequently, on the basis of restriction digestion, methods such
intricate particulars of these protocols presumptively necessitate as Q-RRBS (Yang et al., 2015), MscRRBS (Charlton et al., 2018),
further refinement for single-microbe RNA-seq experiments and scTAM-seq (Bianchi et al., 2022) were developed sequen-
involving microorganisms derived from complex natural com- tially; these methods have increased the throughput by up to
munities. 10,000 cells. Meanwhile, scBS-seq is the first PBAT-based
scDNAm-seq technology developed in 2014 that utilizes the
strategy of two rounds of random priming and analyses of 96
Single-cell genome sequencing (or scDNA-seq)
cells (Smallwood et al., 2014). Subsequently, methods such as
Along with the development of scRNA-seq, several conventional snmC-seq (Luo et al., 2017) and snmC-seq2 (Luo et al., 2018)
bulk DNA sequencing methods have been reconfigured to enable were developed with improved multiplexing strategies and
the analysis of individual cells. Different from scRNA-seq, scDNA- increased throughput to 384 cells. Notably, a challenge in the
seq typically amplifies the 2 copies of the genome in mammalian development of scDNAm-seq technologies is that most methods
cells, which requires sensitive amplification techniques with are based on bisulfite treatment, which causes significant DNA
single molecular detection power and limited biases. scDNA-seq degradation and limited library complexity (Liu et al., 2023b).
measures the DNA molecules within each cell of a given sample Therefore, new scDNAm-seq methods are expected to be
to allow the detection of genetic variants of a certain frequency developed through enzyme-based bisulfite-free approaches, such
and the DNA copy number changes in the assayed cells. In as Cabernet, a recently published method to measure 5mC and its
scDNA-seq, DNA extracted from isolated cells is most frequently oxidative form (5hmC) at a single-base resolution with high

1232 SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 [Link]
genomic coverage (Cao et al., 2023). (Bartosovic et al., 2021; Wu et al., 2021), wherein the enzyme
used for chromatin fragmentation is first fused to protein A/G
Chromatin accessibility (pA/G) and subsequently conjugated to primary antibody. When
Chromatin accessibility is a window into gene regulation, which compared with ChIP-seq-based methods, these methods omit
denotes the activity of regulatory regions and their accessibility chromatin fragmentation before antibody incubation, often
to transcription factors (TFs) (Tsompana and Buck, 2014). The adapting a combinatory indexing strategy to increase the
profiling of chromatin accessibility at the single-cell level is, throughput and automation level (Preissl et al., 2023). In these
therefore, of vital importance to determine cellular heterogeneity methods, no multi-step library preparation is required, which
in gene regulatory programs. Currently, chromatin accessibility minimizes cell loss significantly. Nevertheless, for minimizing pA-
is mainly assessed based on the accessibility of an enzyme such as Tn5 bias under high-salt conditions, a significant challenge is
Tn5 transposase, DNase I, MNase, or GpC methyltransferase encountered when profiling TF binding (Preissl et al., 2023).
(Wen and Tang, 2022). Probably due to the complexity of the
protocol and the intrinsic bias of the enzyme, single-cell DNase- 3D genome organization
seq (Gao et al., 2021) has not been broadly adopted despite being Chromatin architecture refers to the 3D organization of DNA and
developed. In contrast, the Tn5-based ATAC-seq methods are DNA-associated proteins, including topologically associated
readily adaptable for single-cell assay, and various methodologies domains (TADs) and A/B compartments formed therefrom
based on microwell (Chen et al., 2018; Xu et al., 2021b), (Preissl et al., 2023). In 3D genome architecture, distal
nanowell plate (Mezger et al., 2018), microfluidic (Buenrostro et enhancers are placed spatially close to their targeted gene
al., 2015), or droplet (Lareau et al., 2019) have been developed promoters. Single-cell profiling of chromatin architecture is thus
for scATAC-seq. The first reported scATAC-seq method by needed to understand how different genomic regions interact
Buenrostro et al. paved the way for subsequent scATAC-seq with each other in individual cells. High-throughput/resolution
approaches based on droplet, microwell, and pool-split indexing. chromosome conformation capture (Hi-C) is a conventional
A plausible reason for the development of scATAC-seq is that it approach that involves capturing chromatin architecture in bulk
slices the genome and adds adaptor tags to the accessible regions samples. Based on this approach, scHi-C was developed to
simultaneously in a cell, which makes it exemplary for high- capture chromatin architecture at a single-cell level, which is
throughput analysis (Wen and Tang, 2022). Different from also the most widely applied method. In the original study, only
ATAC-seq, the GpC methyltransferase-based method utilizes 10 T helper cells were assayed, with 0.01–0.03 million contacts
exogenous GpC methyltransferase to artificially methylate detected in each cell (Nagano et al., 2013). By omitting the
cytosines within GpC dinucleotides. Considering that GpC biotin-related step, Flyamer et al. (2017) detected a median of
dinucleotide occurs at approximately every 25 bp in the 0.34 million contacts. Subsequently, Tan et al. (2018) increased
mammalian genome, this method provides a higher resolution the number of contacts to a median of 1 million per cell by
and enables the analysis of chromatin states and DNA methyla- employing a Tn5-based whole-genome amplification protocol. In
tion simultaneously within the same single cell (Wen and Tang, this method, Tan et al. (2021) used unindexed Tn5 and profiled
2022). more than 3,000 cells, with approximately 0.4 million contacts
per cell. Meanwhile, combinatorial barcoding methods were
Histone modifications employed in single-cell profiling of chromatin architecture.
Histone modifications are basically posttranslational modifica- Although thousands of cells can be profiled otherwise, the
tions (e.g., acetylation, phosphorylation, and methylation) that resulting data are sparser than those yielded by the scHi-C
involve adding or removing from histone tails so as to modulate a method. To address this point, Mulqueen et al. (2021) employed
specific gene expression (Bannister and Kouzarides, 2011). There a Tn5-based library strategy and increased the number of
are multiple histone activation marks (such as H3K4me1, contacts to approximately 0.1 million per cell. In addition,
H3K4me3, H3K36me3, H3K79me2, H3K9Ac, H3K27Ac, and several rounds of split-pool barcoding were used by Arrastia et al.
H4K16Ac) and repressive marks (such as H3K27me2/3 and in their recent introduction of single-cell split-pool recognition of
H3K9me1/2/3) that display differences in the genomic distribu- interactions by tag extension (scSPRITE) to capture TADs,
tion patterns. To profile histone modifications in individual cells, compartments, and inter-chromosomal interactions (Arrastia et
several methods have been developed, which mainly fall into two al., 2022).
categories. The first group of methods is based on ChIP-seq, such
as Drop-ChIP (Grosselin et al., 2019), scChIP-seq (Rotem et al.,
Single-cell sequencing of other modalities
2015), and itChIP-seq (Ai et al., 2019). The main obstacle in
ChIP-seq-based single-cell sequencing is the low specificity and In addition to the profiling of genome, transcriptome, and
sensitivity of antibody capture and the need for adding cell- epigenome in individual cells, single-cell genomics has branched
specific barcodes before aggregating the cells for immunopreci- out into other modalities like proteins, metabolites, and
pitation (Wen and Tang, 2022). For this reason, itChIP-seq adds ribosomes. First, to comprehend the proteomic constitution of
cell barcodes via Tn5 transposase tagmentation, while scChIP- individual cells, several distinct methods have been exploited,
seq and Drop-ChIP add cell barcodes via MNase digestion and including: (i) mass spectrometry (MS)-based; (ii) antibody-based;
ligation (Wen and Tang, 2022). The second set of methods is (iii) imaging-based; (iv) single-molecule sequencing (Redit et al.,
based on chromatin immunocleavage (e.g., CUT&RUN and 2023). Among these strategies, MS-based methods (e.g.,
CUT&Tag), such as scChIC-seq (Ku et al., 2019), uli-CUT&RUN proteoCHIP) have achieved the most success. Second, to catalog
(Hainer et al., 2019), iscChIC-seq (Ku et al., 2021), ACT-seq the chemical contents of individual cells, an array of novel
(Carter et al., 2019), combinatorial barcoding, targeted chroma- technologies has been developed. For example, scMEP uses high-
tin release (CoBATCH) (Wang et al., 2019a), and scCUT&Tag dimensional antibody-based methods to quantify proteins

[Link] SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 1233
regulating metabolic pathways for the characterization of the modalities to preserve data integrity and analytical efficiency—a
metabolic regulome of individual cells (Hartmann et al., 2021). principle that can be considered as an essential one of multi-
Third, to understand mitochondrial DNA in individual cells, omics.
methods such as mtscATAC-seq (Lareau et al., 2023; Lareau et Multi-omics methods have enhanced our understanding of cell
al., 2021a) have been developed to detect chromatin accessibility state transitions, gene expression profiles, and regulatory net-
and, thereby, mitochondrial genotypes. works in normal tissues and for diseases across species. Further
applications of single-cell multi-omics have been elaborated in
some previous reviews (Baysoy et al., 2023; Vandereyken et al.,
Chapter 3. Single-cell multi-omics technologies
2023). In this section, we have first addressed the various
strategies known for modality separation, which is a key
Introduction to single-cell multi-omics
distinguishing feature between multi-omics and single-omics
Recent advances in single-omics technologies have ushered in a methods. We have then highlighted representative methodolo-
new era of single-cell analysis that has facilitated detailed gies to illustrate the progression and state-of-the-art advance-
examination of cellular diversity across multiple biological ments in multi-omics. In the discussion, we have presented a
layers—from genomics and epigenomics to transcriptomics, comprehensive technical timeline on the combined detection and
translatomics, and proteomics (Figure 1). This evolving field separation of modalities developed over the past decade.
has substantially deepened our current understanding of cellular
heterogeneity and developmental trajectories in health and
Modality separation in multi-omics
disease. To elucidate the complex interplay between these layers,
concurrently profiling multiple molecular modalities within the Over the past decade, multi-omics methodologies have advanced
same cell is crucial. Recent years have witnessed the emergence into systems that incorporate multiple strategies to distinguish
and refinement of single-cell multi-omics approaches that among modalities. An effective multi-omics approach ensures the
integrate genomics and transcriptomics, which have now maximum efficiency of each modality while minimizing signal
expanded to include epigenomics, capturing DNA methylation, loss and contamination or interference. Other beneficial factors
chromatin accessibility, histone modifications, and chromatin include compatibility with standard laboratory setups, cost, and
organization. These approaches also extend to proteomics, hands-on time. Broadly speaking, the evolution of multi-omics
epigenomics, and metabolomics, which provide a holistic view techniques has been guided through the advancements made
of cellular functions and the central dogma. with single-omics methods. For example, the recent development
Single-cell multi-omics profiles have revealed new insights into of the CUT&Tag method represents an advancement over the
the identification of sub-clusters, cellular heterogeneity, and the conventional single-cell ChIP-seq by efficiently labeling modifica-
impact of genomic and epigenomic variation on biological tion sites using Tn5 barcoding (Bartosovic et al., 2021; Wu et al.,
functions. Principally, multi-omics studies aim to detect co- 2021). This advanced method leverages the simplicity of
localization or co-occurrence of omic features within individual segregating barcoded tags from other modalities, thereby forming
cells or across cell types and lineages to provide a basis for the basis for many chromatin-associated protein bi-omics
inferring causal relationships between chromosomal information methods that employ a Tn5 conjugated antibody recognition
and molecular manifestations, including mRNA, ncRNA, and framework. However, multi-omics is more than a mere aggrega-
proteins. Integrating bioinformatics analysis to construct trajec- tion of single-omics techniques. Here, each step is meticulously
tories introduces a temporal aspect to these data, which can optimized; for instance, (i) it involves a robust cell carrier
enhance our understanding of cellular dynamics. Moreover, the platform to enhance efficiency and throughput; (ii) each modality
advent of spatial omics has augmented our knowledge of cellular is distinctly isolated to prevent cross-contamination and signal
distribution within tissues. Single-cell multi-omics technologies loss; (iii) a precise and convenient protocol has been established
are particularly advantageous for measuring limited samples, for amplification and library construction; (iv) there is an
enabling comprehensive multi-dimensional analysis without unbiased data acquisition process and the availability of
needing sample division for separate assays, which potentially appropriate downstream analysis tools. Multi-omics technologies
reduces the overall costs. leverage the platforms, protocols, and analytical techniques of
Single-omics methods have laid the groundwork for multi- singleomics. However, it is the challenge and the necessity of
omics approaches. The transcriptome plays a central role in modality separation that fundamentally distinguishes multi-
linking genotype to phenotype as a crucial intermediary between omics. We have categorized the modality separation strategies
the genome and proteome. As such, transcriptomics, being one of into three: physical separation, in silico separation, and
the most mature and earliest-developed omics, offers multiple enzymatic conversion (Figure 3). Moreover, we have delineated
strategies and has been established as a cornerstone in the amplification strategies into separated and co-amplification
evolution of multi-omics methodologies. Multi-omics techniques, approaches.
which are rooted in single-omics methods, retain the major
distinct procedure, such as the initial cell dissociation and Physical separation
collection via cell capture platforms, including microfluidics, Physical separation techniques partition target molecules into
mouth-pipetting, flow cytometry, or sophisticated labeling distinct pools for individual library construction. A fundamental
strategies like split-and-pool barcoding (Figure 2). Subsequently, method involves extracting a fraction of a cell lysate for
the information from various modalities is transformed into personalized profiling. This strategy serves as a cornerstone for
features that allow sequencing through diverse strategies and current transcriptome and translatome bi-omics methods, which
deciphering through data processing and integration. Crucially, rely on direct lysate distribution (Hu et al., 2016b; Zou et al.,
multi-omics approaches underscore the necessity of separating 2022). Despite potentially losing half the signal, physical

1234 SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 [Link]
Figure 2. Workflow for single-cell multi-omics detection. Cells containing multiple omics information are collected through different platforms similar to that by single-omics
methods. After the modalities are distinguished through different strategies, libraries containing sequence information and barcodes are sequenced, and the information is
processed with pipelines to separate and integrate multiple omics. Different colors indicate distinguished molecules, sequences, or features in individual graphs.

profiling as well; for example, sci-CAR amplifies cDNA and

fragmented DNA from divided lysate portions (Cao et al., 2018).
For enhanced separation, magnetic beads are used to isolate
specific molecules. For example, oligo(dT) beads are used to
extract poly-A RNA from lysates for use in downstream single-
cell RNA-seq protocols. In SCTG, genomic DNA is purified with
AMPure XP beads or ethanol, which may incur additional losses
(Li et al., 2015). Simultaneous high-throughput ATAC and RNA
expression with sequencing (SHARE-seq) combines combinator-
ial indexing with bead-based isolation of cDNA to improve
throughput (Ma et al., 2020b). Bead separation has been
integrated into advanced tri-omics methods such as single-cell
chromatin accessibility, RNA barcoding, DNA methylation
sequencing (scChaRM-seq), single-cell nucleosome occupancy,
methylome and RNA expression sequencing (scNOMeRe-seq),
and single-nucleus methylcytosine, chromatin accessibility, and
transcriptome sequencing (snmCAT-seq) for mRNA isolation,
thereby highlighting the technique’s utility and popularity (Luo
et al., 2022; Wang et al., 2021d; Yan et al., 2021).
Nuclei separation, which involves a two-step lysis, segregates
intact nuclei from cytosolic components. After gentle cell lysis,
nuclei are extracted by following different protocols. One
pioneering multi-omics method employs microfluidics to parti-
tion nuclei for DNA retrieval (Han et al., 2014). High-speed
centrifugation is also an effective means of isolating nuclei, as
demonstrated in scTrio-seq (Bian et al., 2018; Hou et al., 2016).
Plate-based centrifugation has increased throughput for isola-
Figure 3. Different classes of modality-separation strategies. Strategies are classified tion, as observed with scCAT-seq or direct nuclear tagmentation
into three groups: physical separation, in silico separation, and enzyme conversion. and RNA sequencing (DNTR-seq) (Qu et al., 2019; Zachariadis et
Each group includes multiple methods, as described in the main text and selectively
drawn. Different colors and shapes indicate different molecules or elements in each
al., 2020). In addition, beads can facilitate nucleus trapping to
method. Specifically, solid balls indicate enzymes or molecules; the green box indicates form visible clumps for easy manipulation. In the simultaneous
barcode position (in different sequences); the red box indicates Tn5 insertion i5/i7 isolation of genomic DNA and total RNA sequencing (SIDR-seq),
sequence; the yellow or blue box indicates enzyme-cutting sites. antibody-conjugated beads have been used for nuclear precipita-
tion (Han et al., 2018a), while scSET-seq employed Concanavalin
separation can efficiently isolate molecules and maintain the A conjugated paramagnetic beads to bind specific membrane
integrity of labile entities, such as RNA and proteins, under glycoproteins (Sun et al., 2021a). A recent comprehensive study
various lysis conditions. This separation process can occur post- conducted on colorectal cancer employed carboxylic acid beads

[Link] SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 1235
for profiling genomic DNA and mRNA bi-omics (Zhou et al., barcodes. Primers or adaptors are often employed to append
2020). Such a bead-assisted nuclear separation method has been barcodes for targeting molecules during RT or PCR amplification.
incorporated into LiMCA—single-cell transcriptomics plus 3D Barcode separation offers several advantages, as follows: (i) it
genome methodology (Wu et al., 2024a). Nuclei separation obviates the need for additional separation steps, as barcodes are
effectively minimizes cross-contamination between nuclear and integrated during profiling; (ii) it facilitates high throughput by
cytosolic analytes but may result in the loss of molecular combining modality and cell barcodes; (iii) it minimizes signal
complexity, which affects components such as nuclear mRNA loss; (iv) it distinguishes between physically similar omics, such
and mitochondrial DNA. as the various histone modifications. Tn5 transposases, which
The aforementioned strategies aim to separate modalities cleave and tag double-stranded DNA or RNA-cDNA hybrids with
before amplification or, in the initial stages, to ensure complete adaptors, are useful for barcode addition (Penkov et al., 2023).
differentiation in subsequent analyses. However, these ap- In summary, each modality separation strategy has distinct
proaches often entail lower throughput, require more time and applications. Physical separation typically coincides with sepa-
resources, and may cause some signal loss. An enhancement to rate amplification, whereas enzyme conversion and in silico
this approach is implementing physical separation post-amplifi- separation are generally employed before co-amplification. Some
cation. For this reason, DR-seq is devised to co-amplify cDNA and methods generate individual libraries for each modality to
gDNA, followed by partitioning the amplicons into distinct DNA optimize the sequencing depth, while others construct a mixed
and RNA libraries (Dey et al., 2015). Although cDNA can be library to streamline procedures and reduce the processing time.
tagged with specific T7 adaptors, there remains a risk of cDNA A single method may integrate multiple strategies, especially in
contamination in the gDNA library. SNARE-seq employs a splint tri-omics technologies. Although transcriptome profiling has
oligo complementary to the Tn5 adaptor sequence post- utilized all of these three strategies, specific combinations such as
tagmentation, which enables the simultaneous capture of DNA methylation and chromatin accessibility are predominantly
chromatin accessibility and transcriptome. This method allows associated with a single strategy, emphasizing its unique
unique cell barcoding within each droplet, which helps stream- advantage. Thus, the choice of separation strategy and timing
line the process (Chen et al., 2019). is dictated by the modalities involved, and these strategies
continue to evolve to better accommodate diverse modalities.
Enzyme conversion
Enzyme conversion leverages enzyme-driven chemical modifica-
Development of multi-omics
tions to distinguish or label different modalities. Commonly, this
process involves the use of a DNA methylase, such as [Link], to Over the past decade, single-cell multi-omics methods have
modify GpC sites in open chromatin regions that are not bound to undergone significant evolution, progressing from their initial,
nucleosomes, resulting in the conversion to methylated cytosines low-throughput versions that covered a limited range of
(Kelly et al., 2012). Subsequently, unmethylated cytosines are modalities to sophisticated, comprehensive systems that incor-
converted into uracil through bisulfite conversion. This method porate a variety of strategies and encompass nearly all modalities.
allows for the simultaneous profiling of naturally methylated CpG We have compiled here a selection of these methods, ranging
sites across the genome and enzyme-converted GpC sites within from 2014 to 2023, in Table S2. Notably, there has been a
accessible chromatin. Techniques such as ScTrio-seq2, scCOOL- marked acceleration in the development of new methodologies,
seq, NMT-seq, and scNOMe-seq have adopted and refined this particularly since 2021. The integration of modalities, including
approach to create a unified set of methodologies (Bian et al., the genome, epigenome, transcriptome, and other cytoplasmic
2018; Clark et al., 2018; Guo et al., 2017; Pott, 2017). Although analytes, into the multi-omics toolkit has progressively expanded
this strategy tends to favor cytosine-rich loci and provides a during this period.
higher coverage around transcription start sites, it can be
integrated with RNA-seq. For example, cnmCAT-seq incorpo- Genomics and epigenomics
rates 5′-methyl-dCTP during cDNA synthesis in RT, thereby The Central Dogma underscores the flow of genetic information
protecting cDNA from bisulfite damage and eliminating the need from DNA to RNA, highlighting a key objective of multi-omics, i.
for separate mRNA isolation (Luo et al., 2022). e., exploring the connection between genomic or epigenomic
Another enzyme conversion strategy, as exemplified by Paired- variation and transcriptional output. G&T-seq and DR-seq
seq and Paired-tag, employs restriction enzymes to differentiate exemplify two primary methodologies that have distinct
libraries post-amplification (Zhu et al., 2019; Zhu et al., 2021). In approaches (Dey et al., 2015; Macaulay et al., 2015). For
this process, following split-and-pool barcoding of single cells, instance, G&T-seq employs oligo-dT-covered beads to capture
cDNA and DNA are tagged with distinct adaptor sequences. The full-length mRNA for profiling via Smart-like protocols, while
pre-amplified mixture is then partitioned and selectively cleaved gDNA is processed using methods such as MDA or PicoPlex,
with NotI or SbfI. Strategies based on enzyme conversion heavily whichever is applicable. Conversely, DR-seq co-amplifies cDNA
depend on the specific performance of enzymes and represent and gDNA to minimize sample loss, although it possibly
areas primed for further exploration and development. compromises full-length mRNA detection and introduces poten-
tial cDNA interference in DNA signals. TARGET-seq, like DR-seq,
Post-sequencing separation employs co-amplification for targeted mutation detection along-
Another significant category of the separation method involves side cDNA and gDNA primers (Rodriguez-Meira et al., 2019). Sci-
either co-amplification or copool of different modalities, which L3 increases throughput by co-amplifying DNA and RNA with
distinguishes them after sequencing. In some instances, mod- distinct barcodes and combinatorial indexing (Yin et al., 2019).
alities can be directly identified from the sequence. More ScONE-seq facilitates a one-tube reaction for genomics and
commonly, modalities requiring indirect identification utilize transcriptomics with Tn5 insertion post-RT (Yu et al., 2023a),

1236 SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 [Link]
enabling separate sequencing of DNA and RNA libraries at involves the use of DNA adenine methyltransferase (Dam) to
optimal depths. methylate adenines at protein-associated GATC sites, marking
Following the widespread utilization of genomics, studies DNA sites’ interaction with labeled proteins. ScDam&T-seq and
integrating epigenomics have emerged. While genomics con- EpiDamID are designed to cleave these methylated GATC sites in
centrates on individual variations or mutations, epigenetics single cells, coupled with IVT-based amplification (Markodimi-
highlights cell differentiation and gene expression regulation. traki et al., 2020; Rang et al., 2022). DamID relies on in vivo
The first epigenetic multi-omics methods primarily focused on transfection and reaction of Dam fusion proteins; hence, the
DNA methylation. In 2016, MT-seq, scTrio-seq, and M&T-seq, results often reflect cumulative interactions or residence time
based on RRBS or genome-wide BS-seq, followed a separation rather than immediate protein binding states.
strategy similar to that of G&T-seq, thereby providing a Finally, the relatively underdeveloped yet crucial aspect of
foundational framework for DNA methylation multi-omics epigenomics lies in understanding the 3D genome. Hi-C, a widely
(Angermueller et al., 2016; Hou et al., 2016; Hu et al., utilized sequencing method, is based on spatial proximity ligation
2016b). Smart-RRBS, a recent method, involves the application of DNA fragments for profiling chromatin architecture (Lieber-
of tens to thousands of cells without explicit nucleic acid man-Aiden et al., 2009). The initial 3D genome-related multi-
extraction (Gu et al., 2021). Newer methods using enzyme- omics approaches were developed in combination with DNA
based methylation site conversion or third-generation sequen- methylation. ScMethyl-HiC and snm3C-seq detect bisulfite-
cing reduce DNA loss to achieve higher coverage, implying converted methylated sites along with ligated fragments from
potential multi-omics applications (Cao et al., 2023; Sun et al., different genomic loci (Lee et al., 2019; Li et al., 2019). Despite
2021b). potential contact reductions due to bisulfite conversion, these
The earliest chromatin accessibility multi-omics did not stem methods have displayed correlations between methylated CpG
from existing single-omics methods such as DNase-seq, MNase- loci and higher-order chromatin structures, particularly com-
seq, or ATAC-seq (Buenrostro et al., 2015; Lai et al., 2018; Song partments. The first transcriptome and 3D genome method,
et al., 2011). Enzyme conversion using GpC methylase was first named HiRES (Hi-C and RNA-seq employed simultaneously), was
employed in scCOOL-seq and scNOMe-seq for profiling open reported recently (Liu et al., 2023f). HiRES employs barcode
chromatin coupled with DNA methylation (Guo et al., 2017; separation of DNA and cDNA, followed by co-amplification of
Pott, 2017). ScNMT-seq physically isolated mRNA from nuclei, both ligated genomic DNA and cDNA in a single-tube reaction.
subsequently profiled via scRNA-seq, single-cell DNA methyla- Several methods quickly followed this report, combining various
tion, and chromatin accessibility, ultimately becoming one of the strategies such as nuclei separation, barcode separation, and
earliest tri-omics methods, followed by methods such as combinatorial indexing (Luo et al., 2023; Wu et al., 2024a; Zhou
scNOMeRe-seq, scChaRM-seq, scCAT-seq, and snmCAT-seq et al., 2023). Benchmarks for these methods are essential for
(Clark et al., 2018; Luo et al., 2022; Wang et al., 2021d; Yan further evaluation. Considering the integral role of the 3D
et al., 2021) (Liu et al., 2019). Sci-CAR, the initial ATAC-based genome in reflecting chromatin dynamics, it is likely intercon-
method, combined direct splitting of cDNA and DNA fragments nected with other omics aspects. Therefore, new methods that
with combinatorial indexing of pooled cells (Cao et al., 2018) to integrate the 3D genome with other omics are eagerly
profile thousands of single cells into a single experiment. Since anticipated.
2018, ATAC-related Tn5 fragmentation has been widely adopted Our discussion primarily focuses on nuclear DNA, which differs
in chromatin accessibility-associated methods owing to its from cytoplasmic DNA, including extrachromosomal DNA
convenience and reliability, such as SHARE-seq (Ma et al., (ecDNA) and organelle DNA, such as mitochondrial DNA
2020b), SNARE-seq (Chen et al., 2019), Paired-seq, and in situ (mtDNA), which are often excluded from nuclei separation.
sequencing hetero RNA-DNA-hybrid after assay for transposase- Nonetheless, some studies have incorporated the detection of
accessible chromatin-sequencing (ISSAAC-seq) (Xu et al., these DNA types. In PHAGE-ATAC, phage-mediated tagmenta-
2022b). Chromatin accessibility has become a frequently tion enabled the detection of mtDNA mutations (Fiskin et al.,
involved omics, second only to transcriptome (Cheng et al., 2022); mtscATAC-seq adapted a modified 10x ATAC chromium
2021a; Qu et al., 2019; Xing et al., 2020). system for mtDNA detection, which traces thousands of cancer
Histone modifications and TFs, targeted by Tn5-conjugated cells into clones and sub-clones (Lareau et al., 2023). ScEC&T-seq
antibody fragmentation, were initially reported in bulk methods physically separated DNA, allowing ecDNA to be identified by its
and then quickly adapted to single-cell analysis (Kaya-Okur et al., circular feature (Chamorro González et al., 2023); scGTP-seq
2019). Methods such as CoTECH, Paired-Tag, and scSET-seq, similarly captured ecDNA by using a single-molecule real-time
which employ insertion strategies similar to ATAC-related sequencing of long fragments amplified through transposon
techniques and incorporate transcriptomics, have also been insertion (SMOOTH-seq) like processes (Chang et al., 2023a).
developed (Sun et al., 2021a; Xiong et al., 2021; Zhu et al., Both methods confirmed a high correlation between oncogene
2021). Alternatively, scPCOR-seq uniquely uses MNase con- expression and ecDNA formation.
jugated to protein A for histone site localization (Pan et al., CRISPR-mediated gene editing, when combined with single-
2022). Since 2020, the simultaneous profiling of multiple cell approaches, significantly enhances our understanding of
chromatin-associated proteins within the same cells has become genomic element functions. CRISPR-based methods such as
feasible (Gopalan et al., 2021; Janssens et al., 2022a; Lochs et al., CRISP-seq, CRISPR droplet sequencing (CROP-seq), and Perturb-
2024; Meers et al., 2023; Stuart et al., 2022; Tedesco et al., seq co-profile transcriptomics, are often implemented on plates or
2022; Yeung et al., 2023). This strategy has also been combined microfluidics, which enable genome-wide screening (Datlinger et
with epitope detection, as in scCUT&Tag-pro, wherein epitopes al., 2017; Dixit et al., 2016; Jaitin et al., 2016). Recent
were profiled using oligo-conjugated antibodies alongside advancements involve integrating gRNAs with ATAC-seq to
scCUT&Tag data (Zhang et al., 2022a). Another strategy explore the effects of gene perturbation on chromatin accessi-

[Link] SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 1237
bility. Perturb-ATAC and Spear-ATAC allow single-cell ATAC- single-cell RNA and immunodetection (RAID-seq), and inCITE-
seq on microfluidics platforms, which enriches reverse-tran- seq (Chung et al., 2021; Gerlach et al., 2019; Reimegård et al.,
scribed gRNA with barcoded primers (Pierce et al., 2021; Rubin 2021). However, this strategy is limited to proteins for which
et al., 2019). CRISPR-sciATAC increases throughput with efficient antibodies are available, thereby restricting its scope. A
combinatorial indexed ATAC-seq and two-round PCR enrich- recent method, single-cell simultaneous transcriptome and
ment of gRNA (Liscovitch-Brauer et al., 2021). Expanded proteome (scSTAP), tackles this issue by dividing oocyte lysate
CRISPR-compatible cellular indexing of transcriptomes and (like R2-lite) and profiling RNA and proteins separately through
epitopes by sequencing (ECCITE-seq) incorporates protein in- MATQ-seq and MS-based techniques (Jiang et al., 2023c).
formation, facilitating precise sub-clustering of immune cells Metabolomics focuses on small molecules, such as lipids or
based on the receptor type (Mimitou et al., 2019). sugars, that play critical roles in signaling pathways and energy
processes. Metabolic profiling typically relies on MS, which may
Post-transcription omics cause irreversible damage to the molecules. Research in the field
Multi-omics technologies are mainly centered around transcrip- of single-cell metabolomics and its combination with proteomics
tomics, not only because of the extensive development and is emerging (Li et al., 2021b; Shrestha, 2020). One of these
sophistication of transcriptomic methodologies but also because approaches involves the use of oligo-EG conjugated microwells to
of its central role in the flow of genetic information. However, extract metabolites from adsorbed proteins for separate MS
RNA profiling encompasses more than just mRNA quantifica- analysis. However, metabolomics methods remain on the fringe
tion. For example, full-length RNA sequencing provides addi- of multi-omics, and there is a need to develop more innovative
tional insights into alternative splicing and allelic expression; combinations.
these aspects may be lost in methods that focus solely on 3′- or 5′-
enriched cDNA. Another critical aspect is the posttranscriptional
New era in single-cell multi-omics: tech choices and further
modifications of RNA, which play crucial roles in mRNA
applications
regulation, RNA metabolism, and translation activity. One
notable modification in mammalian mRNA and lncRNA is Overall, currently, known single-cell multi-omics can be
m6A (Sun et al., 2019). Recently, sn-m6A-CT has been developed characterized based on several key features: (i) the majority of
based on a 10x Chromium platform to simultaneously profile the current methods focus on a combination of transcriptomes with
full transcriptome and m6A methylomes (Hamashima et al., other genomic modalities such as chromatin accessibility, DNA
2023). These m6A methylomes were found to effectively methylation, chromatin-associated proteins, and 3D genome; (ii)
differentiate cell types and exhibit variations in several pluripo- a range of previously unexplored modalities have recently been
tent marker genes. incorporated into multimodal assays; (iii) there is a rapid increase
The coupling of transcription and translation has raised in the number of new methods; (iv) established methods are
increasing attention recently. Ribo-seq enables the profiling of progressively being applied to address biological problems and
ribosome-bound mRNA both at the population and single-cell then contribute to the construction of tissue atlases. Conse-
levels (VanInsberghe et al., 2021; Xiong et al., 2022). However, quently, the advancement in the development of new mono-
detecting single-cell ribosome-bound mRNA poses challenges modal profiling workflows and strategies for modality separation
owing to its scarcity and potential loss. T&T-seq and R2-lite have are propelling multi-omics research forward.
successfully separated transcriptome and ribosome-bound trans- Ideally, a mature multimodal method should offer comparable
latome in mammalian oocytes or embryos (Hu et al., 2022; Zou efficiency across each single-ome, feature a user-friendly proto-
et al., 2022). T&T-seq captures ribosome-mRNA complexes with col, and be cost-effective with the availability of accessible
A-site beads, whereas R2-lite directly splits the lysate for downstream analysis tools. Certain combinations, such as
individual profiling. Notably, R2-lite identified pre-produced integrating 3D genomes with chromatin accessibility and histone
mRNA in oocytes functional in early embryos, thereby poten- modifications or tri-omics involving transcriptome, translatome,
tially initiating zygotic gene activation. Hence, single-cell and proteome, are currently unachievable. For combinations
translatome profiling can help uncover post-transcriptional that involve multiple methodological variations, systematic
regulation across cell types by following a method that applies comparison of these methods and their standardization is
to cells of standard size. essential for widespread, large-scale applications. Here, we
compared several strategies with throughput, time, costs,
Post-translation omics availability, and core instruments (Table 1). For laboratories
Cytosolic analytes, which encompass a diverse array of molecules with limited single-cell experience, fewer cell diversity, and
in the cytoplasm, present unique challenges and opportunities for simple biological models, some easy and cheap methods can be
single-cell analysis. Unlike nucleic acids, proteins cannot be sufficient, such as those including physical separation methods or
directly amplified or sequenced, which poses significant hurdles mouth-pipetting platforms. Microfluidics, combinatorial barcod-
in single-cell proteomics. Nevertheless, proteins can be indirectly ing, or in silico optimization can benefit a throughput design that
profiled by using oligo-conjugated antibodies. Techniques such serves huge atlas projects or complex systems. As mentioned
as cellular indexing of transcriptomes and epitopes by sequencing earlier, multi-omics methods greatly depend on modality
(CITE-seq), as well as RNA expression and protein sequencing separation and data quality. Modality contamination or exogen-
assay (REAP-seq), have successfully helped profile surface ous contamination should be particularly avoided because
epitopes with dozens of oligo-conjugated antibodies, thereby multiple modality reactions extend over the entire method
facilitating transcriptome-like analysis. The extensions of this period, which introduces additional potential risks. Some
approach to intracellular proteins have been achieved using methods share similar limitations, such as the excessive loss in
methods such as Single-Cell Protein and RNA Co-profiling, direct separation or RNA-to-DNA contamination in in silico

1238 SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 [Link]
Table 1. Comparison of different strategies, including modality separation and single-cell platforms
Strategy Throughput Time Cost Availability Core instruments
Direct lysate distribution Medium Fast Low Easy Pipette
Bead separation Low Medium Medium Medium Molecules binding beads
Nuclei separation Low Medium Medium Medium Nuclei binding beads
Enzyme conversion Usually High Fast Medium Easy Converting or cutting enzymes
In silico separation Usually High Medium Low Required well-designed algorithms Good PC
Mouthpipette Low Slow Low Required skillful person Mouthpipette tube
Plate sorting Medium Medium High for instrument Required FACS in clean room FACS for single-cell plating
Microfluidics High Medium High for instrument High requirement Microfluidics
Combinatorial barcoding High Fast Medium Required primers Plates and barcode primers

methods such as DR-seq. As there is no perfect method Spatial omics technology, which is emerging in the post-Human
established yet, some trade-offs should be made when selecting Genome Project era, aligns with the historical trajectory of
a method that better serves the project. genetic molecular detection methods and introduces a novel
It has been shown in the past that when compared with single framework for unraveling the biological genetic code (Marx,
dimensional methods, multi-omics methods mainly study an 2021).
internal correlation among diverse cell types of different The method employed in this study to identify the relative
modalities. In the long past, scientists applied DNA/RNA bi- spatial arrangements and expressions of tissues, cells, and
omics to study mutation and clonal expansion with crucial gene biomolecules for deciphering the mechanisms of biological events
switching in leukemia (Zachariadis et al., 2020). These correla- is termed “spatial omics technology” (Larsson et al., 2021).
tions were further verified as causality by methods, including Although various multi-omics approaches, such as spatial
gene editing (Bhattarai-Kline et al., 2022). As multi-omics are genomics and spatial proteomics, have emerged, the current
centralized with transcriptomes, RNA information can help focus of spatial omics technology predominantly centers on
cluster basic cell types in complex tissues or provide an output for measuring and analyzing transcripts (Tang, 2021). This spatial
gene regulation of the genome or epigenome. RNA dynamics detection technique, at the transcriptional level, has a long-
reflect a cell lineage within the development process, which standing history. In its early stages, direct detection of the
produces a temporal perspective to study other omics. The localization of nucleic acid and molecules within tissues and cells
potential interactions of different epigenomes can be seen in bi- posed a significant challenge. A clever solution involved the use
epiomics methods; for instance, DNA methylation can mold of artificially introduced fluorescent probes for in situ hybridiza-
chromatin architecture to present a diversity of neural cells (Lee tion of nucleic acids within tissues (Femino et al., 1998). By
et al., 2019; Tian et al., 2023b). A widespread application of detecting the fluorescent probe, it acts as a surrogate for directly
post-transcription methods, including translatome, proteome, or identifying nucleic acid molecules. In essence, fluorescent
metabolome, remains short. These methods can be gradually imaging can be employed to observe the spatial positions and
unlocked with the development of new methods or improvement abundance of nucleic acid expression within cell or tissue
in efficiency. Alternative strategies developed in the future can sections. This methodology has been recognized as spatial in situ
facilitate the establishment of a comprehensive system and hybridization (SISH) technology (Figure 4). Diverging from the
setting up industry benchmarks. Moreover, computational principles of hybridization imaging technology, which entail
methods have evolved to precisely distinguish different modalities chemically labeling base fragments for identification, the
and even integrate different single-ome data into “multi-omics advancement and maturation of techniques such as PCR,
integrated” data. With the rapid evolution of single-cell multi- ligation-based methods, and synthesis-based sequencing have
omics technologies, we can expect more insightful discoveries facilitated the identification of individual bases in nucleic acid
associated with diverse fundamental biological processes such as sequences with robust signals (Koboldt et al., 2013; Markham,
development, aging, and diseases. 1993). For instance, in the case of in situ sequencing (ISS) (Ke et
al., 2013), following the binding of target RNA with padlock
probes, in situ amplification is performed, followed by ligation-
Chapter 4. Spatial genomics technologies
based sequencing to ultimately determine the spatial character-
istics of nucleic acid information. This technique, performing in
Introduction of spatial genomics
situ base sequencing to glean genetic molecular information, is
“The Human Genome Project” stands as a pivotal moment in recognized as spatial ISS (SISS) technology.
contemporary biology, which initiated humanity’s systematic With the accomplishment of “The Human Genome Project,”
capability to decipher the genetic code of living organisms the introduction of Roche’s first 454 next-generation sequencers
(International Human Genome Sequencing Consortium, 2004). in 2005, and Illumina’s release of the Genetic Analyzer 2 in
Positioned as a transitional milestone in the evolution of genetic 2007, the landscape of biological sequencing transformed low-
molecular detection techniques, it has progressed from scrutiniz- throughput to high-throughput NGS (Metzker, 2010). In this
ing individual molecules to high-throughput molecular synchro- high-throughput sequencing paradigm, all nucleic acids are
nization detection (Amaral et al., 2023). This progression also extracted by subjecting them to biochemical reactions on the
signifies the shift from a single-genome era to the current age and entire tissue. Subsequently, they are amalgamated in a single
phase of multi-omics integration (Vandereyken et al., 2023). tube for sequencing, alignment, and assembly, which marks a

[Link] SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 1239
Figure 4. Timeline of spatial transcriptomics and spatial multi-omics methods.

1240 SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 [Link]
significant shift from the conventional workflow that involves (Cho et al., 2021), provides single-cell precision through the
sequential hybridizing and sequencing of individual nucleic development of sequencing chips.
acids. This shift also invited novel perspectives for spatial omics In summary, spatial omics technology can be classified into
technology, giving rise to spatial in situ microdissection (SISM) two primary categories based on the distinct inherent principles
technology. Through high-precision microdissection techniques, for achieving spatial resolution and then subdivided into four
specific regions of tissue sections are physically separated, and subcategories: (i) Spatial Imaging: SISH and SISS; (ii) Spatial
the dissected tissues undergo high-throughput sequencing of Sequencing: SISM and SISB. The rapid advancement of these
mixed nucleic acid libraries (Emmert-Buck et al., 1996). This spatial omics technologies has significantly expedited scientific
approach helps associate each sequencing result with a unique research in the fields of developmental biology, neuroscience,
spatial position during the reconstruction of spatial locations. pathology, botany, and other cutting-edge research domains
The early implementations of spatial SISM encompass Tomo-seq (Moses and Pachter, 2022). These innovations offer unique
(Junker et al., 2014) and commercially available Digital Spatial biological insights into a range of previously unexplored life
Profiling (DSP) (Merritt et al., 2020) techniques. SISM involves phenomena (Table S3).
the initial physical cutting of selected regions, followed by the
assignment of spatial location markers. However, this process
Spatial technologies
limits spatial resolution because of the constraints introduced by
the physical cutting often performed on cell populations (Bressan SISH technology
et al., 2023). Although single-cell cutting is achievable, it smFISH (Femino et al., 1998) is regarded as the classical SISH; it
remains exceptionally challenging. Consequently, the enhance- involves the combination of individual RNA molecules in tissues
ment of spatial resolution has posed a limitation in the evolution with multiple fluorescently labeled probes. This technique allows
of spatial omics technology. To overcome the constraints the capture of images, facilitating the detection of RNA spatial
associated with physical operations, it is imperative to challenge localization and expressions. The fundamental logic underlying
the conventional logic of slicing first and then assigning spatial SISH can be summarized in two key steps: (i) Oligonucleotide
information. A breakthrough emerged in this direction by pre- Probes Binding to Targeted RNA Fragments by using specific
labeling tissue cells and subjecting them to subsequent biochem- oligonucleotide probes that bind to target RNA fragments and (ii)
ical analyses, thereby mitigating the limitations of physical Capturing Images of Fluorescent Groups on Probes to detect
operations. Furthermore, the potential for independent labeling signals and positions of RNA molecules. However, SISH
of individual cells or even nucleic acids within cells promised an encounters challenges with respect to signal-to-noise ratio and
escalation in the resolution from multicellular to single-cell or detection sensitivity, which necessitates optimization of these
subcellular levels. The foundation of conventional sequencing core steps with various solutions. However, enhancing the signal
technology can be referred for this innovative strategy. In single- value and sensitivity of RNA molecule detection can be achieved
cell sequencing approaches, barcode probes can pre-label each by increasing the number of probes binding to each mRNA (Raj
nucleic acid within individual cells, followed by biochemical et al., 2008). Moreover, branch amplification techniques, such as
reactions and sequencing for individual cells (Kolodziejczyk et al., branch DNA “double Z” probes (Player et al., 2001), hybridiza-
2015; Li and Wang, 2021; Sankaran et al., 2022). The tion chain reaction (HCR) (Shah et al., 2016), and rolling circle
advantages of barcode technology include the following: (i) amplification (RCA) reaction can help enhance the detection
advanced labeling of cells and nucleic acids, which circumvents signal following probe-target hybridization (Wu et al., 2018).
the limitations of cutting first and then assigning spatial ClampFISH (Rouhanifard et al., 2019) employs click chemistry
information. (ii) A vast array of barcode probes that theoretically for probe end hybridization and signal amplification is achieved
exceed the number of nucleic acids in the tissues, ensuring a through successive rounds of hybridization and click reactions in
unique barcode for each nucleic acid. (iii) Integration of barcodes this method. The detection signal can also be amplified by
with polyT to form capture probes that are specific for the 3′ applying the overall cycling smFISH strategy (Codeluppi et al.,
polyA tails of mRNA, thereby enabling unbiased transcriptome to 2018) and through the incorporation of primer-exchange
capture instead of targeting. (iv) Implantation of barcode probes reactions (Kishi et al., 2019) (Figure 5).
on diverse carriers, including common magnetic beads and The incorporation of spectral combinatorial labeling strategies
modified glass slide surfaces. By leveraging the maturity of and high-resolution, super-resolution microscopy in imaging can
barcode technology and the progress in spatial omics, Ståhl et al. significantly augment the detection capabilities and sensitivity,
(2016) pioneered the implantation of barcodes on glass surfaces thereby offering a solution for optical signal congestion. In
in 2016. This technique allowed unbiased capture of nucleic seqFISH (Lubeck and Cai, 2012), each gene is designated a
acids within covered tissues, which, when followed by in situ RT, unique color post-hybridization, and upon probe stripping, the
can generate an abundance of amplified sequences for high- subsequent rounds of hybridization and colorimetric detection
throughput NGS. This groundbreaking method, known as spatial ensue. Theoretically, by employing 4 colors and 8 rounds of
in situ barcode (SISB) technology, achieved high-throughput amplification, the entire genome can be encoded, yielding
transcription detection within tissue sections based on NGS in the 48=65,536 potential codes. RNA SPOT expands the color palette
spatial dimension. Thereafter, diverging from the implantation of to 12 “pseudo-color panels,” and seqFISH+ (Eng et al., 2019)
barcode probes on flat surfaces, subsequent innovative techni- further extends it to 60 colors. Through the amalgamation of
ques emerged. For example, Slide-seq (Rodriques et al., 2019) images and error-correcting hybridization, precise imaging of
combines barcode sequences on magnetic beads, allows determi- 10,000 genes can be achieved using this approach.
nistic barcoding in tissue for spatial omics sequencing (DBiT-seq) Another conventional method that employs spectral combi-
(Zhang et al., 2020b), and introduces barcodes through fluidic natorial barcoding is multiplexed error-robust FISH (MERFISH)
channels, while Stereo-seq (Chen et al., 2022a) and Seq-Scope (Chen et al., 2015). It characterizes genes as “1” and “0” by

[Link] SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 1241
Figure 5. Overview of the four main spatial transcriptomics methods.

employing hybridization and colorimetric detection. Subse- accurately detect and categorize each base, thereby completing
quently, the genes are encoded into a binary form through the sequencing process. Consequently, the crux of SISS lies in the
multiple rounds of hybridization. Error-correcting codes are in-situ bulk amplification of nucleic acid fragments. PCR
implemented using a minimum Hamming distance, facilitating technology forms the fundamental logic for nucleic acid
the detection of 10,000 genes. Integration with the expansion amplification, and the initial proposed solution for cloning and
microscopy further enhances the overall density of detectable amplifying DNA on glass microscope slides using PCR reactions is
RNA, as with MERFISH (Wang et al., 2018a). Similarly, EASI- Polony technology (Mitra and Church, 1999). In 2003, building
FISH achieves 3D detection of brain slices with a thickness of upon Polony technology, Church’s group (Mitra et al., 2003)
300 μm by combining the use of hydrogel and amplifying the introduced the first-generation Fluorescent ISS (FISSEQ) technol-
detection signal via the HCR (Wang et al., 2021c). ogy, enabling ISS of amplified DNA sequences from in situ PCR
In addition to the conventional MERFISH and seqFISH and demonstrating its application potential in tissues like
techniques, spatial in situ detection methods inspired by the bacteria and mice. In 2014, the second-generation FISSEQ
principles of smFISH have emerged, each offering unique technology emerged (Lee et al., 2014). Differing from its
technical characteristics and application scenarios. For instance, predecessor, this iteration directly captures non-targeted RNA
EEL FISH (Borm et al., 2023) utilizes electrophoresis to transfer molecules within tissue sections, reverse-transcribes them into
RNA from tissue sections onto a capture surface for data cDNA, amplifies them through RCA, and employs phi29
acquisition, offering distinct advantages with respect to signal- polymerase. This process generates numerous continuous
to-noise ratio and speed. The par-seqFISH (Dar et al., 2021) amplified copies fixed in the three-dimensional gel space,
method allows the recording of the gene expression and spatial allowing for subcellular resolution ISS and showcasing potent
background in micro assemblies at a single-cell level with application capabilities in tissues such as mouse brains, embryos,
molecular resolution. SISH has experienced rapid enhancements and Drosophila embryos (Lee et al., 2015). Another notable SISS
in signal-to-noise ratio, detection capabilities, sensitivity, and is ISS, introduced by Ke et al. (2013) in 2013. ISS utilizes padlock
application scenarios with further development. Nonetheless, the DNA probes to bind to target nucleic acid sequences, followed by
intricacies of multiple hybridizations and limitations in detection gap filling and ligation to form a complete circular structure.
capabilities due to optical congestion continue to impede Amplification is then performed, and connection-based sequen-
significant technological breakthroughs. cing is employed to achieve spatial ISS. FISSEQ and ISS
technologies have established the foundational framework for
SISS technology SISS, encompassing tissue and nucleic acid fixation, nucleic acid
The distinction between SISH and SISS lies in the necessity for a amplification, and ISS. These innovations have paved the way for
more robust nucleic acid signaling in sequencing so as to subsequent SISS techniques with enhanced efficiency, resolution,

1242 SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 [Link]
and specificity. tion method involves conducting high-throughput sequencing
Built upon ISS technology, the HybISS (Gyllborg et al., 2020) on the entire composite library, as opposed to capturing precise in
method integrates barcoded padlock probes (PLPs) and RCA situ images of individual nucleic acid sequences. This strategy
approaches, utilizing bridged probes for sequence-by-hybridiza- mitigates the constraints associated with multiple hybridizations
tion chemistry and highly multiplexed detection, in contrast to and optical congestion to provide the benefits of batch processing
the sequence-by-ligation scheme employed in ISS. This modifica- and high-throughput detection. However, it is apparent that
tion facilitates higher-throughput spatial targeted sequencing of delineating distinct spatial dimensions within the mixed sequen-
larger tissue areas. In contrast, IISS (Tang et al., 2023c) alters cing library has emerged as a new bottleneck and, hence, a
the probe binding scheme through combinatorial probe anchor- constraint in the process.
ing chemistry (icPAL) and adopts a two-base barcode strategy, In SISM, the imparting and restoration of spatial information
enhancing sequencing quality, signal intensity, and specificity. for tissues and cells before and after sequencing constitute
Spatially Resolved and signal-diluted Next-generation Targeted fundamental elements of the technique. This is achieved through
sequencing (SPRINTseq) (Chang et al., 2023b) employs synthetic physical partitioning or chemically labeling selected regions of
sequencing by synthesis to expedite the reaction alongside a interest (ROIs). Subsequently, the isolated libraries undergo
microfluidic flow cell and a novel padlock probe hybridization either bulk sequencing or single-cell sequencing, which allows
design. This significantly reduces the time needed for ISS, the straightforward detection of spatial heterogeneity between
boosting detection throughput. Remarkably, SPRINTseq enables ROIs. LCM (Emmert-Buck et al., 1996), a widely employed
the detection of 142 million transcripts in 450,000 cells in <2 microdissection technology, employs laser irradiation on the
days, which showcases a substantial advancement in technol- selected regions by fusing the artificial membrane with the
ogy. The enduring principles of ISS underscore the remarkable underlying cells. This process involves region segmentation,
progress achieved in the development of these innovative recovery into a container, and subsequent sequencing of different
technologies. libraries. Tomo-seq (Junker et al., 2014) further enhances spatial
The integration of SISS with hydrogel technology is emerging resolution by slicing the entire zebrafish embryo into 50–100
as a notable trend. The incorporation of hydrogel in tissue sections, extracting RNA from each section for individual
clearing serves to diminish optical interferences from the tissue sequencing and thereby achieving a relatively coarse-resolution
background. STARmap (Wang et al., 2018b) exemplified this 3D ST detection. Building on the principles of LCM and Tomo-seq,
trend, excelling in achieving high-resolution ISS of three- higher-resolution spatial omics technologies have been devel-
dimensional tissues. In this methodology, STARmap/STARmap oped, such as Geo-seq (Peng et al., 2016), immuno-LCM-RNAseq
PLUS initially uses pairs of DNA probes to bind the target mRNA (Zhang et al., 2022b)—which incorporates immunofluores-
within the tissues, resulting in amplified DNA nanoballs (DNBs) cence-guided laser cutting—STRP-seq (Schede et al., 2021)—
anchored within the three-dimensional hydrogel space. Subse- which involves the cutting of adjacent tissue sections at different
quently, tissue clearing is executed by eliminating proteins and angles—and GaST-seq (Giolai et al., 2019)—which is specifically
lipids, paving the way for in situ sequencing, and enabling the designed for plant research.
detection of over 1,000 genes within the cleared tissues (Shi et Another application of ROI involves labeling with photoactive
al., 2023; Zeng et al., 2023b). ExSeq (Alon et al., 2021) uses the groups within the tissues, followed by biochemical reactions to
expandability of hydrogel to secure tissue nucleic acids within the assign spatial identifications (IDs) for sequencing detection.
hydrogel. Similar to FISSEQ, ExSeq is capable of nontargeted Notable techniques in this category include TIVA-tag (Lovatt et
sequencing through the amplification of RNA into cDNA al., 2014), PIC (Honda et al., 2021), and ZipSeq (Hu et al., 2020).
nanoballs. This method showcases the prospect of achieving Among these, DSP (Merritt et al., 2020) technology has achieved
subcellular resolution in diverse tissues, including mouse commercial success.
hippocampal neurons, visual cortex, and human breast cancer. In the context of SISM, spatial positioning information is
Building upon the imaging methodologies of SISH and SISS, applied to tissues, and high-throughput measurements are
enhancements have been implemented in hybridization princi- conducted on the entire library. However, the constraints
ples, combinatorial labeling, error correction schemes, and encountered in physical sectioning operations considerably limit
various other aspects. When coupled with advanced microscopy enhancement in spatial resolution. This limitation poses a
and hydrogel technology, these techniques have progressively challenge when conducting large-scale studies at a single-cell
elevated throughput and sensitivity for spatial in situ detection or subcellular level.
across the transcriptome. The authors even demonstrated a
significant potential in true 3D spatial omics. However, SISB technology
challenges continue to persist, such as in the experimental Microdissection technology involves the segmenting of various
intricacies of multi-hybridization imaging schemes, optical regions of tissues. The potential to differentiate each mRNA
congestion arising from high-density nucleic acids, and the within a cell and perform high-throughput sequencing presents
inherent low-throughput nature of current sequencing schemes. an opportunity to improve spatial resolution from the cellular
These factors collectively hinder fundamental parameter im- level to the single-cell or even the subcellular level. Spatial
provements, making it challenging to harness the full potential of tagging technologies have played a crucial role in supporting this
second-generation sequencing technologies. concept. In 2016, Lundeberg’s group (Ståhl et al., 2016)
introduced ST technology, a groundbreaking approach that
SISM technology incorporated high-density probes with spatial barcodes and polyT
In the period following the initiation of the human genome sequences on a glass surface for preparing a capture chip. Tissue
project, spatial omics technologies have evolved in tandem with slices were placed on the chip, and treatments were applied to
second-generation sequencing techniques. The primary detec- enable penetration and perforation of cell membrane structures,

[Link] SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 1243
facilitating the targeted capture of the 3′ polyA structures of individual high-resolution base sequencing is performed on each
mRNA within cells by using the chip’s probes. In situ RT was amplification sequence. Over continuous iterations, the mini-
then performed to generate cDNA, which was subsequently mum physical spacing of nucleic acid sequences on Illumina and
dissociated from the glass slide, collected into a container, BGI sequencing chips reaches the nanometer level, with the
amplified by PCR, and subjected to high-throughput sequencing. number of nucleic acid clusters on a single chip reaching
By aligning barcode sequences uniquely associated with mRNA hundreds of millions. The parameters of these chips, whether
in the sequencing results with the known barcode positions physical or optical, have pushed the boundaries of the field to
during implantation, the spatial locations of mRNA can be their maximum. Considering this point, the idea of replacing the
calculated, thereby enabling spatially resolved transcriptome nucleic acid fragments on the sequencing chip with capture
sequencing. The spot spacing of ST technology, which refers to probes consisting of different combinations of barcode and polyT
the distance between the centers of spots, is 200 μm, which gives sequences opens the possibility of capturing and sequencing
a spatial resolution of 200 μm. Following its acquisition by 10x tissue mRNA. Theoretically, this approach can achieve nan-
Genomics and subsequent technical optimization, ST technology ometer-level/subcellular-level spatial resolution omics technolo-
was rebranded as “Visium,” offering an improved resolution of gies. The imminent challenge lies in reaching the technological
100 μm. Although Visium does not achieve an absolute single- limits of the field in the short term. In the realm of technological
cell resolution, considering the presence of multiple cells within progress, Stereo-seq based on BGI sequencing chips and seq-
each spot, it still offers significant technical advantages when Scope (Cho et al., 2021) based on Illumina sequencing chips have
compared with microdissection techniques. In addition, ST gradually emerged around 2022, which aligns with the
technology is the first to successfully assign spatial IDs to mRNA trajectory of technological development and reshaping of spatial
using barcodes, thereby introducing an original concept for omics technologies. The distinction lies in the composition of the
subsequent spatial omics technologies and laying the foundation Stereo-seq chip’s surface, which comprises matrix-arranged spot
for the current popular spatial omics methodologies (Cheng et al., points that leverage BGI sequencing’s patented DNBs to carry
2023a). spatial barcode probes, which achieve a consistent resolution of
The essence of barcode-based spatial omics technologies lies in 500 nm. Registering and segmenting single-stranded DNA
the method of assigning distinct barcodes to individual mRNA (ssDNA)-stained tissue images has accomplished genuine sin-
molecules. In ST technology, different barcode probe clusters are gle-cell resolution in spatial sequencing. In addition, owing to the
physically implanted at specific distances from each other on a expensive dimensions of BGI sequencing chips’ reaction chip, the
glass slide to ensure minimal interference. An alternative Stereo-seq spatial chip can span larger centimeter-level sizes,
approach involves carrying different barcode probes on separable reaching up to 13 cm×13 cm, which facilitates spatial omics
carrier units, which are then assembled to form a chip with the detection across entire sections of monkey brains. Both two
carrier as the minimum spatial resolution. For example, Slide-seq single-cell resolution SISB methods involve comparisons of
(Rodriques et al., 2019) and Slide-seqV2 (Stickels et al., 2021) diverse barcodes within mixed libraries to ascertain mRNA
have implemented this concept by using magnetic beads, which positions, necessitating two rounds of sequencing for each chip
are widely being used presently, whose spatial resolution is and subsequently escalating the experimental costs. Piexl-seq (Fu
determined by the diameter of the magnetic beads, which is et al., 2022) introduces an innovative solution wherein barcode
typically 10 μm. In contrast, high-definition STs (HDST) (Vicko- transfer on the capture chip is accomplished by following a
vic et al., 2019) use smaller magnetic beads that are placed in stamp-like method. This novel approach enables the acquisition
microwells to discriminate spatial barcodes so as to achieve a of several dozen original spatial omics chips in a single
finer spatial resolution of 3 μm. However, due to limited detection sequencing run, thereby ultimately reducing the overall opera-
signals, only the regions formed by multiple magnetic beads can tions cost.
be used as the minimum analysis unit without attaining single- Sequencing-based spatial omics technologies such as SISM and
cell resolution. Microfluidic devices have demonstrated notable SISB offer notable technical merits amid the evolution of high-
advantages in the distribution of spatial barcodes. In the case of throughput sequencing technology. For instance, they have
DBiT-seq (Zhang et al., 2020b), a microfluidic device is employed, excelled in achieving unbiased whole transcriptome detection,
permitting separate liquid addition in the x and y directions. This and their enhanced throughput and user-friendly operation have
capability facilitates the addition of x-axis and y-axis coordinates propelled their widespread adoption in the spatial omics
to mRNA within tissue slices to accomplish spatial ID transposi- technology market. However, as previously mentioned, sequen-
tion. Despite these advancements, the spatial resolution is cing chip-based spatial omics technologies have, to an extent,
constrained by the limitations of the microfluidic channel size, encountered constraints in the field, which have posed chal-
and it has not yet attained the level of single-cell resolution. lenges for significant short-term enhancements to key technical
The evolving trend in barcode technology is focused on reading parameters. Naturally, the development of technology hinges on
and editing nucleic acid sequences with higher spatial resolution the imperatives of scientific research. The existing strengths of
accuracy. This process involves equipping them with unique STs suffice to cater to the demands of a majority of research
sequence spatial barcodes for planar capture and RNA sequen- endeavors, thereby exhibiting robust application capabilities
cing. Unlike the development of barcode assignment schemes across domains such as neuroscience, developmental biology,
from the ground up in previous technologies, high-throughput pathology, and plant biology (Tian et al., 2023a). In the
sequencing technologies have been naturally progressing in this immediate future, the discernible technological trajectory in-
direction over several decades. High-throughput sequencing volves transitioning from single STs to multi-omics, encompass-
schemes entail biochemically attaching numerous cDNA ampli- ing genomics, epigenomics, proteomics, and much more. The
fication sequences with specific physical spacing onto the surface pivotal focus is thus on achieving simultaneous detection of
of a sequencing chip (Reuter et al., 2015). Subsequently, multiple omics within the same tissue section.

1244 SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 [Link]
Spatial multi-omics genomics technologies remains limited. Therefore, future
technological advancements should focus on developing meth-
Spatial multi-omics research spans genomics, epigenomics,
ods that enable the simultaneous detection of the genome and
transcriptomics, proteomics, and metabolomics, as well as
transcriptome.
investigations into transcriptional and translational states
(Vandereyken et al., 2023). Presently, spatial omics technologies
Spatial epigenomics
predominantly center on the transcriptome dimension for
Spatial epigenomics introduces a novel perspective for analyzing
deducing the expression patterns and regulatory mechanisms
the spatial dimension of cellular fate regulation mechanisms, and
in biological phenomena through mRNA abundance analysis. In
its technological development has rapidly progressed in tandem
the context of central dogma, RNA expression is not only the
with transcriptomics technology. Zhuang’s group (Lu et al.,
outcome of upstream gene transcription and epigenetic regula-
2022) employed the MERFISH technique based on multiplexed
tion but also the catalyst for downstream protein translation. The
SISS to capture specific epigenetic modifications on chromatin in
central dogma has evolved into the “spatiotemporal central
situ. The authors labeled the DNA near the modification site with
dogma,” which allows simultaneous characterization of the
a T7 promoter, transcribed the labeled DNA fragments in situ to
biological phenomena at multiple omics levels, encompassing
generate RNA, and detected and characterized the target RNA
temporal and spatial dimensions, allowing for a more compre-
using MERFISH’ multiplexed in situ fluorescence hybridization
hensive understanding of cellular fate regulation mechanisms.
technology, thereby enabling spatial epigenomic detection, called
This evolution significantly enhances our microscopic under-
epigenomic MERFISH. Sequencing-based spatial omics technol-
standing of macroscopic life phenomena. Pioneering spatial
ogies have exhibited enhanced compatibility in omics studies. In
omics technologies are forging a technological ecosystem geared
spatialATAC, Ståhl’s group (Llorens-Bobadilla et al., 2023)
toward concurrent multi-omics detection, with a primary
integrated Tn5 transposase into the tissue clearing step of ST-
emphasis on the transcriptome dimension. These technologies
based RNA capture technology, which has allowed the labeling
exhibit promising prospects, particularly in domains such as
of open chromatin and the restoration of spatial position through
mouse brain research and disease studies (Figure 6).
spatial barcoding, thereby achieving spatial ATAC detection.
Similarly, by leveraging microfluidic channels, Fan’s group
Spatial genomics proposed spatial-ATAC-seq (Deng et al., 2022b) and Spatial
Despite the fact that genome research boasts a lengthy history, CUT&Tag (Deng et al., 2022a) technology. The Spatial CUT&Tag
the technology available for delineating the spatial dimension of involves the introduction of antibodies targeting modified
chromatin states or DNA information within tissues remains proteins into tissue sections, followed by enhancement of pA-
nascent. Hi-M (Cardozo Gizzi et al., 2019), which is grounded in Tn5 transposase binding with secondary antibodies. Next,
multiplexed imaging, represents a method that can concurrently adapters are inserted into the antibody recognition sites of DNA
discern chromosomal organization and transcriptional informa- with histone marks, and the sequences with spatial barcodes are
tion within individual nuclei. However, due to constraints in in situ connected to the adapters through bilateral channels.
imaging technology and hybridization capabilities, Hi-M en- Subsequent sequencing then generates a genome-wide map for
counters limitations in detecting tissues and DNA loci. Through the spatial analysis of histone modifications.
the optimization of multiple hybridization schemes, SISH has Microfluidic indexing-based spatial assay for transposase-
markedly enhanced the detection of cell numbers and genomic accessible chromatin and RNA-sequencing (MISAR-seq) adopted
loci. DNA-MERFISH technology (Sibai et al., 2020), which similar strategies for spatially resolved joint profiling of chroma-
employs sequential hybridization, provides direct visualization of tin accessibility and gene expression (Jiang et al., 2023a). In
chromatin structural domains and compartments, unveils conclusion, epigenetic regulation stands as a crucial facet for
interchromosomal interactions, and facilitates simultaneous comprehending the mechanisms underlying microscale cellular
imaging of over 1,000 gene loci and nascent transcripts. DNA fate regulation and macroscopic biological events such as
seqFISH+ (Takei et al., 2021) allows imaging of 3,660 development and disease. Spatial epigenetic regulation technol-
chromosome loci in mouse embryonic stem cells. Similar to ogy has gradually emerged and garnered attention. However, the
transcriptomics technology, in situ imaging-based techniques, simultaneous detection of the epigenome and transcriptome
such as SISH and SISS, primarily characterize targeted loci, encounters technical challenges, hindering the rapid develop-
while sequencing-based methods capture unbiased whole-tissue ment of simultaneous multi-omics epigenetic technologies. This
DNA information. Leveraging the foundation of Slide-seq, area thus remains a significant avenue for future advancements
Chen’s group (Zhao et al., 2022) pioneered Slide-DNA-seq by in multi-omics technology.
using barcode and capture sequence modifications on densely
packed beads to achieve an unbiased capture of DNA within the Spatial proteomics
tissue surface. Following sequencing, spatial genomic informa- In the realm of spatial omics technology, the essential
tion has become accessible and applicable in the study of human prerequisite for accurately pinpointing biomolecules in space lies
cancer samples, uncovering clonally specific genetic alterations in their specific capture in situ or an unbiased capture, followed
across diverse genomes. Although slide-DNA-seq can be by reconstruction using the carried positional IDs. Guided by
integrated with adjacent sections for multi-omics characteriza- these principles, proteins can be detected in situ through the
tion, it presently falls short of achieving simultaneous multi- specific binding of antigen-antibody interactions, which can then
omics sequencing on the same tissue section. In summary, facilitate targeted spatial detection. Alternatively, proteins can be
spatial genomics technology offers distinctive insights into quantified and sequenced by linking oligonucleotide barcodes
developmental biology, human diseases, and various research with antibodies, thus leveraging the CITE-seq principle. Owing to
domains. However, the availability of widely applicable spatial the stability of proteins and the compatibility of experimental

[Link] SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 1245
Figure 6. Timeline of the application of spatial omics methods in the fields of development, neuroscience, pathology, and plants.

1246 SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 [Link]
procedures, achieving simultaneous detection of proteins and Comprehensive comparison of spatial omics technologies
transcriptomes on the same tissue section is relatively straight-
forward. This capability has enabled true spatial dual-omics Various spatial omics techniques are rapidly evolving, necessitat-
technology encompassing proteomics and transcriptomics and ing a thorough evaluation of their respective strengths and
provided crucial technical support for the rapid development and limitations for targeted applications in specific contexts. Compar-
widespread application of spatial proteomics technology. ison across categories of these techniques generally encompasses
The method for detecting targeted proteins in situ, following spatial resolution, detection efficiency, signal diffusion, and tissue
specific binding with antibodies, relies heavily on the area.
execution of the final imaging step. Therefore, this approach (1) Spatial resolution. Imaging-based spatial omics methodol-
facilitates the direct spatial dimension detection of targeted ogies, as exemplified by MERFISH (Moffitt et al., 2016) and
proteins in imaging-based ST technologies. This method is STARmap (Wang et al., 2018b), facilitate the detection of
frequently employed in conjunction with the existing RNA individual molecules within cells at the subcellular resolution
detection methods to integrate stable protein detection level. This enables single-cell resolution through spatial segmen-
schemes. In imaging mass cytometry (Schulz et al., 2018), tation of single cells by using cell staining techniques.
RNA is amplified in situ by using the RNAScope technique, Differentially, sequencing-based spatial omics approaches gen-
and metal chelate antibodies are subsequently introduced to erally involve the capture of nucleic acid molecules via probes,
bind to the targeted proteins. Following staining, in situ with subsequent spatial position reconstruction achieved
imaging is conducted to achieve spatial dual-omics detection. through barcoding. Consequently, the spatial resolution of the
CAD-HCR (Liu et al., 2022c) technology amplifies RNA latter is contingent upon the probe density on the capture
signals in situ through the HCR, and fluorescence imaging is substrates. Presently, the majority of sequencing-based spatial
performed after achieving a specific binding between anti- omics methodologies function at a multicellular resolution, with
bodies and proteins. STARmap PLUS technology, an extension each capture unit encompassing multiple cells, as exemplified by
of STARmap’s RNA in situ detection, allows imaging of Visium (Suo et al., 2022) and DBiT-seq (Liu et al., 2020b).
proteins after the in-situ binding of primary and secondary Nonetheless, certain techniques exhibit probe densities that can
antibodies, thereby enabling spatial protein detection. This attain subcellular resolution and spatial single-cell precision can
approach has been instrumental in the determination of be achieved via cell image segmentation, as demonstrated in
spatial-specific expression of amyloid-beta and tau protein Stereo-seq (Chen et al., 2022a), Seq-Scope (Cho et al., 2021), and
plaques, which are critical in the progression of Alzheimer’s Pixel-seq (Fu et al., 2022).
disease (AD) (Zeng et al., 2023b). (2) Detection efficiency. The evaluation criteria for detection
This approach involves pre-binding antibodies to nucleotides efficiency currently vary between imaging-based and sequen-
carrying unique barcode sequences and polyA tails, followed by cing-based spatial omics techniques. smFISH is frequently
specific binding to antigens. This process allows the capture of employed to compare the detection efficiency owing to its high
antigens alongside mRNA sequences by the polyT capture sensitivity, wherein the average number of transcripts detected
probes. High-throughput sequencing of the sequences carrying per gene in each cell or region is assessed. Imaging-based spatial
barcodes enables the reconstruction of their spatial positions. omics techniques generally exhibit a higher detection efficiency;
This CITE-seq-based strategy proves more suitable for sequen- for instance, MERFISH achieves an efficiency of approximately
cing-based spatial omics technology. Stereo-CITE-seq technology 95% (Moffitt and Zhuang, 2016). Conversely, sequencing-based
(Liao et al., 2023), built upon Stereo-seq, employs antibodies spatial omics techniques often exhibit reduced detection effi-
with antibody-derived tags (ADTs) sequences for targeted ciency, attributable to the type of capture and library sequencing
proteins. Capturing the ADTs allows synchronous quantification methodologies used. For example, STs detect only 6.9% of the
of transcriptomes and spatial protein detection. Other techniques UMIs per unit area when compared with smFISH (Ståhl et al.,
employing the ADT strategy include DBiT-seq based on micro- 2016). Imaging-based spatial omics techniques typically target
fluidic channels (Zhang et al., 2020b), SM-Omics and SPOTS specific nucleic acid sequences, with optical constraints encoun-
based on ST technology, and Spatial-CITE-seq based on Slide-seq tered during in situ imaging or sequencing, which limits the
(Liu et al., 2023e). In addition, an atomic effort has been made to number of targeted nucleic acids that can be detected in a single
profile spatially resolved translatomics (e.g., RIBOmap (Zeng et experiment. Consequently, optimizing these optical limitations
al., 2023a)), metabolomics (e.g., SEAM (Yuan et al., 2021), and remains a primary focus toward enhancing the efficacy of
scSpaMet (Hu et al., 2023)) at the single-cell level. Overall, imaging-based spatial omics techniques. Presently, both MER-
sequencing-based spatial omics technologies exhibit stronger FISH (Xia et al., 2019) and seqFISH+ (Eng et al., 2019) can
compatibility for capturing proteins and transcriptomes. Syn- detect 10,000 genes. Theoretically, sequencing-based spatial
chronous dual-omics analysis thus holds significant potential in omics techniques have the potential to capture the entirety of
pathological research. mRNA within cells by using polyT probes. Therefore, the primary
In broader terms, spatial multi-omics technology is still in its objective in evaluating these techniques is to calculate and
early phases, and achieving simultaneous spatial detection of enhance the total number of reads captured per unit area or in a
multiple omics on a single tissue section is a formidable single cell. For instance, in an adult mouse coronal hemibrain
challenge. The current workaround involves segregating the section, Stereo-seq can detect 1,910 UMIs and 792 genes per cell.
detection of different omics on adjacent tissue sections, (3) Signal diffusion. The spatial detection accuracy of RNA
followed by integration of the data for analysis. In the future, significantly influences the precision of detection outcomes. This
the pursuit of true spatial multi-omics on a single tissue section parameter is extensively examined in sequencing-based spatial
is bound to emerge as a crucial direction for technological omics methodologies. A contributing factor to this scrutiny is the
advancement. phenomenon wherein mRNA migrates a short distance post-

[Link] SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 1247
enzymatic reaction -permeabilization and is subsequently cap- development of axolotl brain tissues, Wei et al. (2022) identified a
tured. The extent of this diffusion is contingent upon tissue type particular cell type that can generate three distinct subpopula-
and the duration of the enzymatic reaction. For instance, in the tions, each occupying different spatial positions. Similarly, Zhang
comparative analysis of the Slc17a7 marker gene within the et al. (2023a) characterized three clusters of cells with subtly
olfactory bulb, both Slide-seq V1.5 and PIXEL-seq exhibit varying transcriptomes, which were mapped to specific regions in
relatively robust performance with respect to diffusion. Con- the human embryonic limb. The spatial relationships offer a
versely, when evaluating the Pmel marker gene in ocular tissues, critical perspective in cancer research, particularly in the
Stereo-seq demonstrates a superior performance (You et al., elucidation of the advantageous interactions between cancer
2024). cells and surrounding immune cells in cancer cell migration (Ji et
(4) Tissue area. Owing to the inherent limitations in detection al., 2020; Karras et al., 2022). Notably, Erickson et al. (2022)
methodologies, imaging-based spatial omics techniques often identified distinct clonal patterns within tumors and adjacent
face challenges in executing large-scale detection within a single benign tissues by analyzing spatial CNVs. This approach provides
experiment. These methods are typically constrained in that they a framework for understanding the spatially continuous transi-
can detect a range of only a few hundred to a few thousand tion from benign to malignant tissues. In the domain of
genes. Conversely, sequencing-based spatial omics techniques neuroscience, spatial omics methodologies offer significant
exhibit greater flexibility. The effective capture area of these advantages for uncovering the distinctive architecture and
techniques is contingent upon the probe arrangement on the functions of brain tissues. For example, these techniques have
capture substrate. For instance, Visium encompasses an area of delineated a spatial atlas of the mouse central nervous system at
6.5 mm×6.5 mm (Denisenko et al., 2024). PIXEL-seq features an molecular resolution (Shi et al., 2023), delineated the cell type
effective capture area of 75 mm×25 mm, and Stereo-seq can specificity within the primary motor cortex of the mouse (Zhang
generate spatial omics data for monkey brain tissues with et al., 2021c), and highlighted the critical role of the spatial
dimensions of 5 cm×3 cm (Chen et al., 2023a). The scalability connectivity network of neurons in determining the specific
and flexibility of these methodologies render them particularly functions of brain regions (Chen et al., 2023a; Fang et al.,
suitable in the investigation of complex tissues and large-scale 2022a). In the realm of plant biology, the presence of cell walls
spatial gene expression patterns. poses challenges for conventional single-cell sequencing techni-
ques. Spatial omics technology and leveraging of tissue section-
ing and tissue clearing methods triumphs over the hurdles posed
Analysis and application
by cell walls (Li et al., 2023b; Xia et al., 2022). This approach
With the emergence and widespread adaptation of spatial omics enables spatial in situ signal capture, presenting breakthrough
technology, the primary emphasis of spatial omics technology in methodology for botanical research. The transcriptional profiles
scientific applications has gradually shifted from technical of upper and lower epidermal cells in Arabidopsis leaves display
implementation to the analysis and processing of output data nuanced differences. For instance, Xia et al. (2022) effectively
(Palla et al., 2022a). However, without an efficient, direct, and identified and differentiated these spatially distinct yet transcrip-
effective analysis approach for handling large volumes of tionally similar cell subtypes by analyzing the differential
generated data, it is difficult to distill core detection indicators expression characteristics. In addition, Du et al. (2023)
from the massive datasets for guiding biological research. This postulated the presence of two types of meristematic-like cell
hindrance can ultimately lead to data wastage and impede pools within secondary vascular tissues and corroborated the
progress in the field of spatial omics research. Therefore, the data hypothesis through multiple methodological approaches. In
analysis stage has evolved into a limiting factor in spatial omics summary, the emergence and development of spatial omics
technology, encompassing aspects such as data segmentation, technology contributes to technological convenience and acces-
database integration, the development of multi-omics analysis sibility to various research fields, thereby fostering advancements
methods, 3D reconstruction algorithms, and much more. The in our current understanding of complex biological systems
introduction and continuous improvement of efficient and (Table S4).
practical analysis tools are poised to bring greater convenience
and novel research methodologies to scientific research.
Chapter 5. Computational challenges
Spatial omics technology has ushered in new insights and
research perspectives across diverse fields, spanning develop- With rapid advancements in scRNA-seq technologies, a flood of
mental biology, neuroscience, pathology, and botany. The computational tools has been developed to handle the massive
foundational step in understanding biology often involves scRNA-seq data. Generally, these tools are developed for data
identifying the cell types and delineating cellular or genetic preprocessing, batch-effect correction, data normalization, dimen-
functional characteristics (Nolan et al., 2023). By providing a sion reduction, feature selection, cell clustering, cell type annota-
spatial dimension perspective, spatial omics offer a means to tion, graphic presentation, and other purposes. For example, Cell
unearth novel cell types (Zeng, 2022). The integration of spatial Ranger was developed for data preprocessing; Harmony (Lake et
features facilitates the establishment of connections between al., 2019) was introduced for batch-effect correction; and Seurat
function and structure. During the developmental process, (Hao et al., 2021) was developed for normalization, dimension
continuous spatial readability can significantly enhance the reduction, feature selection, and cell clustering. Although manual
crosstalk analysis between differentiated cells and their regula- annotation with a predefined set of marker genes is the gold
tion of cell differentiation by the surrounding cellular factors standard method for cell type annotation, a set of tools such as
(Arutyunyan et al., 2023; He et al., 2022). SingleR (Aran et al., 2019), CellAssign (Zhang et al., 2019a), and
This approach facilitates the identification of cell subtypes that Garnett (Pliner et al., 2019) have been developed to annotate cells
operate within distinct spatial locales. For instance, in the automatically to some extent. In addition, some useful tools have

1248 SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 [Link]
been developed to infer cell trajectory (Qiu et al., 2017; Street et variables for downstream analysis (e.g., clustering), such as
al., 2018), cell-cell communication (Browaeys et al., 2020; VEGA (Seninge et al., 2021), scDHA (Tran et al., 2021), and
Cabello-Aguilar et al., 2020; Efremova et al., 2020; Jin et al., scMM (Minoura et al., 2021); (iii) infer single-cell developmental
2021), metabolic state (Wagner et al., 2021), cell-type-specific progress, such as SPADE (Anchang et al., 2016) and DPT
regulon (Ma et al., 2020a), and somatic mutation (Vu et al., (Haghverdi et al., 2016); (iv) integrate multi-omics datasets, such
2019) from scRNA-seq data. as MARIO (Zhu et al., 2023), Seurat (Hao et al., 2021), LIGER
Computational analyses based on the scDNA-seq data mainly (Liu et al., 2020a), GLUE (Cao and Gao, 2022), StabMap
encompass mutation detection, mutation clustering, genotype (Ghazanfar et al., 2024), Cobolt (Gong et al., 2021), and MultiVI
inference, and phylogeny reconstruction. To address this (Ashuach et al., 2023). In addition, computational efforts were
concern, plenty of tools have been developed that aim to: (i) made to (i) introduce graph construction from single-cell spatial
detect SNVs or CNVs, such as SCAN-SNV (Luquette et al., 2019), data, such as Squidpy (Palla et al., 2022b), Athena (Martinelli
SeCNV (Ruohan et al., 2022), and SCOPE (Wang and Jiang, and Rapsomaniki, 2022), and SPEX; and (ii) detect metabolites
2022; Wang et al., 2020a); (ii) perform mutation clustering and from individual cells, such as SpaceM (Rappez et al., 2021).
genotype inference, such as AMC (Yu and Du, 2022), BnpC Here, we have discussed five key computational challenges
(Borgsmüller et al., 2020), and SCClone (Yu et al., 2022); (iii) encountered in single-cell genomics.
phylogeny reconstruction based on either or both of SNVs and
CNVs such as ConDoR (Sashittal et al., 2023), Phertilizer (Weber
From bulk tissues to single cells
et al., 2023), MEDALT (Wang et al., 2021a), BiTSC (Chen et al.,
2022d), and CellPhy (Kozlov et al., 2022). In addition, some Recent advancements in computational deconvolution methods
useful tools have been developed to detect doublets (e.g., have been proposed for estimating cell type abundances in tissue
doubletD (Weber et al., 2021)) or quantify intra-tumor hetero- RNA-seq profiles (Figure 7) for identifying cell type-specific gene
geneity (e.g., Dhaka (Rashid et al., 2021) and rcCAE (Yu et al., expression alterations and detecting rare cell populations. These
2023c)) from scDNA-seq data. methods are categorized based on their methodology, including
Considering the high dimensionality and sparsity of single-cell linear models (Li et al., 2020; Monaco et al., 2019; Peng et al.,
epigenomic data, dimension reduction is a critical step. To 2019b; Shen-Orr et al., 2010; Wang et al., 2019b; Zhong et al.,
address this concern, numerous analytical tools have been 2013), support vector regression (Newman et al., 2015; New-
developed. A popular method is latent semantic indexing, which man et al., 2019), gene set enrichment approaches (Aran et al.,
has been implemented in many tools, such as Signac (Stuart et 2017; Quon et al., 2013; Yoshihara et al., 2013), probabilistic
al., 2021) and ArchR (Granja et al., 2021). In this method, the models (Kang et al., 2019), and deep learning models (Chen et
term frequency-inverse document frequency (TF-IDF) and al., 2022c; de la Fuente et al., 2023; Menden et al., 2020).
singular value decomposition were used to transform the sparse However, various benchmarking efforts have demonstrated that
count matrix. In addition, methods including spectral embedding the performance of these deconvolution methods can be
(e.g., SnapATAC (Fang et al., 2021)), topic modeling (e.g., influenced by factors such as data volume, data transformation,
cisTopic (Bravo González-Blas et al., 2019)), and variational and the composition of the reference matrix. For example,
autoencoders (AEs) (e.g., SCALE (Xiong et al., 2019)) were enrichment-based approaches and probabilistic approaches often
introduced to reduce the dimension by transforming the sparse impose limitations on the number of usable cell types (Im and
count matrix into a cell-cell similarity matrix, thereby defining a Kim, 2023), while deep learning-based methods encounter
set of “topics” for similar features or encoding features via neural challenges in comprehensively simulating crucial rare compo-
network. Owing to the wide application of single-cell chromatin nents (Zhang et al., 2023d). In addition, solid tissues frequently
accessibility assays, most of these tools were developed for this contain closely related cell types, and the gene expression
specific modality, with the possibility to be adapted and applied to correlation among these cell types introduces collinearity,
other epigenomic data. Another critical step in the analysis of thereby showcasing challenges in resolving their relative
single-cell epigenomic data is to identify and characterize cell proportions in bulk data. The collinearity problem can be
type-specific regulatory elements and their targeted genes. For improved by selecting marker genes through support vector
scATAC-seq data, tools such as Signac (Stuart et al., 2021), regression, as demonstrated in Cell-type Identification By
ArchR (Granja et al., 2021), and SnapATAC (Fang et al., 2021) Estimating Relative Subsets Of RNA Transcripts (Newman et
provide the functions to identify cluster-specific peaks and peak- al., 2015) and BSEQ-sc (Baron et al., 2016) or by employing a
to-gene links. For scDNAm-seq data yielded by snmC-seq (Luo et tree-guided procedure that recursively focuses on closely related
al., 2017), sn-m3C-seq or snmCAT-seq (Luo et al., 2022) and cell types, as exemplified in MuSiC (Wang et al., 2019b). Apart
ALLCools (Liu et al., 2021) enables the identification of from reference-free deconvolution methods, such as DECODER
differentially methylated region (DMR) and DMR-to-gene links. (Peng et al., 2019b) and CDSeq (Kang et al., 2019), most
For scHi-C data, SnapHiC (Yu et al., 2021b) has been developed methods rely on prior knowledge of cell types based on either cell
to detect chromatin loops and predict target genes based on type-specific gene expression signatures or scRNA-seq/snRNA-
physical interactions. seq data obtained from the same tissues. When constructing a
With the accumulation of multi-layered modalities, the cell reference atlas, it is therefore important to integrate data
efficient integration of multi-omics datasets poses a huge from the same tissue samples, consider the effect of scRNA-seq
challenge for computational biologists. To address this concern, and snRNA-seq, address data processing, standardize cell type
a variety of algorithms were developed for the following purposes: annotation, and enhance cell type resolution (Maden et al.,
(i) removing batch effects from multiple datasets, such as MNN 2023). Overall, computational deconvolution can provide a
(Haghverdi et al., 2018), Scarnorama (Hie et al., 2019), and convenient and efficient means for analyzing bulk data from a
Conos (Barkas et al., 2019); (ii) extract single-cell level latent single-cell perspective.

[Link] SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 1249
Figure 7. Key computational challenges presented in single-cell genomics analysis.

From a single cell to millions of cells mentioned in some previous reports (Adil et al., 2021; Poirion et
al., 2016), high levels of zero inflation pose a key challenge in
When compared with bulk RNA-seq analysis, scRNA-seq offers analyzing scRNA-seq data. To address these dropout issues, cell
the advantage of exploring cellular processes with higher imputation methods have been developed (Patruno et al., 2021).
resolution. However, scRNA-seq is more vulnerable to various Model-based imputation methods utilize specific statistical models
disturbances, including systematic differences derived from such as the Gamma-Gaussian mixture model (Li and Li, 2018),
experimental procedures, dropout events, sequencing depth, Poisson-Gamma mixture model (Huang et al., 2018), zero-
and others. To address these challenges and conduct data inflated negative binomial model (Risso et al., 2018), and
preprocessing and noise removal, numerous bioinformatics binomial model (Tang et al., 2020) to identify similar cells.
algorithms are specifically dedicated to single-cell analysis. The Smooth-based imputation methods employ smoothing or diffu-
general workflow of single-cell analysis typically involves quality sion techniques for determining gene expression values in cells
control, alignment, feature quantification, normalization, dimen- with similar expression profiles, for example, DrImpute (Gong et
sion reduction, differential expression, and clustering analysis. As al., 2018), Markov affinity-based graph imputation of cells

1250 SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 [Link]
(MAGIC) (van Dijk et al., 2018), and kNN-smoothing (Wagner et methods, including anchor-based CCA and anchor-based reci-
al., 2018). In addition, data reconstruction methods employ low- procal PCA (RPCA).
rank matrix-based approaches (Linderman et al., 2022; Mongia To provide an unbiased guide to the choice of the BEC method,
et al., 2019; Zhang and Zhang, 2018) or deep learning methods benchmarking studies on integration methods have been
(Amodio et al., 2019; Arisdakessian et al., 2019; Deng et al., conducted. In small-scale samples (comprising hundreds or
2019; Eraslan et al., 2019; Lopez et al., 2018) to represent cells thousands of cells), ComBat outperforms nonlinear methods
in a low-dimensional linear space or latent form for imputation. with the quantitative metric k-nearest-neighbor batch-effect test
Recent studies have compared and evaluated systematic bench- (kBET), which evaluates batch-effect removal and dataset
marks of these scRNA-seq imputation methods (Cheng et al., integration (Büttner et al., 2019). For medium-scale samples
2023b; Hou et al., 2020). Although MAGIC and single-cell (hundreds of thousands of cells), Harmony, LIGER, and Seurat v3
analysis via expression recovery (SAVER) (Huang et al., 2018) are best recommended for batch integration (Tran et al., 2020) as
have outperformed other methods (Hou et al., 2020), they they compare four benchmarks across 14 BCE methods (includ-
impute all zeros in the expression matrix, which increases the risk ing kBET, local inverse Simpson’s index (Korsunsky et al., 2019),
of overfitting for imputation (Zheng et al., 2023). Furthermore, average silhouette width (Rousseeuw, 1987), and adjusted rand
the effectiveness of imputation varies across datasets (Cheng et index (Hubert and Arabie, 1985)). However, best practice
al., 2023b). Therefore, it is essential to provide data-driven recommendations vary for different sample scenarios, such as
guidance for scRNA-seq imputation analysis based on the specific BBKNN and Harmony, exhibiting strong performance in mixed
characteristics and challenges presented by each dataset. heterogeneous biodiversity samples (Chen et al., 2021b). In
Advancements in single-cell atlas mapping technology have large-scale samples (millions of cells), deep learning-based BEC
facilitated the expansion of single-cell datasets in terms of both methods such as scVI and scGen perform well, especially in
size and quantity (Angerer et al., 2017), which has helped handling complex integration tasks (Luecken et al., 2022).
establish common cell atlas consortium databases containing Moreover, the runtime efficiency is a crucial consideration for
millions of single cells (Ye et al., 2024a). Consequently, single- large-scale data integration. Overall, benchmarking integrations
cell genomics is powering big data science, thereby introducing such as scIB serve not only to aid researchers in selecting the
new computational challenges and opportunities for data available integration methods but also in assisting developers in
integration at the atlas level. To address the challenges of building more efficient methods (Luecken et al., 2022).
analyzing millions of cells, frameworks have been developed for Besides BEC methods, another crucial data integration issue
single-cell data integration, such as the Python-based Scanpy involves the harmonization and automated labeling of cell types.
(Wolf et al., 2018) and the R-based Seurat v5 (Hao et al., 2024b). For example, scANVI (Xu et al., 2021a) represents a semi-
The major hurdle in achieving a common ontology of cell types supervised method that extends the scVI model to enable
for downstream analysis in single-cell atlas-level integration is automated annotation of cell types and states. CellTypist
the presence of batch effects (Eisenstein, 2020; Lähnemann et al., (Domínguez Conde et al., 2022) employs logistic regression
2020) arising out of variations ranging from technical to models with stochastic gradient descent learning to identify cell
biological factors (Luecken et al., 2022). Various batch effect types and provides built-in and pre-trained models for different
correction (BEC) methods have been developed based on different tissues. As an upgraded version of CellTypist, CellHint (Xu et al.,
models, ranging from early linear decomposition models to the 2023a) focuses on cell type harmonization. By utilizing a
more recent approaches such as mutual nearest neighbors predictive clustering tree, CellHint can efficiently evaluate cell-
(MNN)-based and deep learning models (Ryu et al., 2023). Bayes- cell similarities and harmonize cell annotations across multiple
based Combat (Johnson et al., 2007) and PCA-based Harmony datasets.
(Korsunsky et al., 2019) are the popular linear decomposition Integrating large-scale single-cell data into a common frame-
models. A typical class of MNN-based BEC methods, such as work is therefore essential for constructing a unified cell census
MNNCorrect (Haghverdi et al., 2018), fastMNN (Haghverdi et al., from a single cell to millions of cells. This comprehensive resource
2018), and Scanorama (Hie et al., 2019), use MNN between not only enhances our current understanding of human biology
batches as anchors. In addition, methods like scMerge (Lin et al., at the cellular level but also facilitates the exploration of diverse
2019) and sMNN (Yang et al., 2021) initially cluster each batch cellular functions and interactions in the human body.
and then identify MNN between the clusters. Furthermore, MNN
has been combined with non-negative matrix fractalization in
From single-omics to multi-omics and spatial omics
methods such as Linked Inference of Genomic Experimental
Relationships (LIGER) (Welch et al., 2018), K-nearest neighbors Recent advancements in single-cell high-throughput technolo-
(KNN) in BBKNN (Polański et al., 2020), and deep learning gies have enabled biologists to catalog various cell types across
models in deepMNN (Zou et al., 2021). In addition, deep multiple omics layers, such as chromatin accessibility (scATAC-
learning-based BEC methods utilize a diverse model framework, seq (Cusanovich et al., 2018)) and transcriptome (scRNA-seq),
with MMD-ResNet (Shekhar et al., 2016) employing a residual together providing a more comprehensive perspective to uncover
neural network (ResNet (He et al., 2016)); SAUCIE (Amodio et the underlying mechanism for cell diversity as well as the
al., 2019) using a deep neural network with an AE workflow; changes occurring during disease development or progression.
scGen (Lotfollahi et al., 2019) implementing a deep neural Unlike single-omics data, which provides information limited to a
network with variational AE (VAE); scVI (Lopez et al., 2018) single molecular dimension, multi-omics data offers a multi-
combining VAE with a probabilistic approach. In addition, Seurat faceted view of cell identity by integrating information from
continuously updates its integration workflows. Seurat v2 uses multiple layers of cellular function. The integration creates a
canonical correlation analysis (CCA) for BEC, and a higher complex interplay of dependencies and relationships, thereby
versions (Seurat v3 to v5) are used in “anchor-based” BEC presenting unique challenges for analysis, such as data integra-

[Link] SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 1251
tion, data heterogeneity, dimensionality reduction, computa- (Ma and Zhou, 2022), Cell2location (Kleshchevnikov et al.,
tional demands, and data interpretation. 2022), and Tangram (Biancalani et al., 2021). Another
Emerging single-cell protocols such as SHARE-seq (Kim et al., integrative application is to reconstruct the spatial association
2023b) and CITE-seq (Stoeckius et al., 2017) enable simulta- of cells in scRNA-seq data by using ST data as references. For
neous profiling of multiple omics, offering a foundational basis for example, scSpace (Qian et al., 2023) can identify spatial cell
integrating multi-omics data at a single-cell resolution. To utilize subpopulations by using a transfer learning model and then
and combine these datasets of multiple types, various novel reconstruct cells onto a pseudo-space.
computational methods that are both effective and precise have In conclusion, computational methods for multimodal integra-
been proposed. Overall, these integration methods can be broadly tion are fast gaining popularity in single-cell analyses and
classified into three types: horizontal, vertical, and diagonal data provide innovative insights into complicated cellular systems and
integration (Athaya et al., 2023), depending on cell pairing and biological processes.
anchor information. Vertical integration aims to integrate
multimodal single-cell data within the same cells through the
From a single species to cross-species
correlation of information from distinct modalities. The
“weighted-nearest neighbor” (WNN) method (Hao et al., 2021) Cell types are evolutionary units with the potential for
is an example of vertical integration that applies an unsupervised independent evolutionary variations (Arendt et al., 2016). The
framework to learn cell-specific modality weights and construct a increasing number of single-cell organism-level datasets from
WNN graph for integrative analysis. WNN analysis has been different species creates opportunities to explore evolutionary
leveraged to pair single-cell transcriptomes and ATAC-seq relationships between cell types across species. Cross-species
profiles, which benefits the multimodal definition of cell identity analysis provides valuable insights into the origin and evolution
as well as the identification of putative regulatory programs. of cell types, shedding light on species-specific expression
Despite the emergence of simultaneous assays, different omics are patterns. Furthermore, many studies use model organisms, such
typically measured independently, which gives rise to unpaired as mice, zebrafish, and fruit flies, in the investigation of
data, requiring horizontal and diagonal data integration. fundamental biological processes and disease mechanisms so as
Horizontal data integration relies on a set of shared features to obtain valuable information about how cell types execute their
(such as genes) to link data from different modalities, including functions in both healthy and diseased bodies, thereby contribut-
CCA (Butler et al., 2018) and LIGER (Welch et al., 2018). CCA ing to our understanding of biological processes across diverse
has been explored to integrate scRNA-seq and scATAC-seq organisms. Specifically, cross-species analysis, such as in the
datasets by transferring chromatin accessibility to gene activity context of complex diseases like IgA nephropathy (IgAN) (Chen
scores, while LIGER integrates gene expression and DNA et al., 2024b), offers a powerful tool in the identification and
methylation by assigning the methylated regions to the nearest understanding of shared disease mechanisms across species and
gene. Furthermore, many multimodal deep learning (MDL) pinpointing potential therapeutic targets.
approaches adopt the diagonal data integration strategy. For MetaNeighbor (Crow et al., 2018) works based on the notion
each high-dimensional data type, MDL can generate a low- that cells of the same type share greater numbers of similar gene
dimensional representation of single cells and convert it into a expression patterns than cells from different types; it has been
common latent space for integration by building on model applied to systematically assess the transcriptional similarity
architectures such as variational AEs (VAE), AE, convolutional between species. The results of MetaNeighbor among humans,
neural network (CNN), and recurrent neural network. Some past mice, and monkeys (Han et al., 2020; Qu et al., 2022) imply that
studies integrated more than modality pairs, as exemplified by the cell type similarity in orthologous gene expression transcends
GLUE (Cao and Gao, 2022). As a deep learning framework mammalian species divergence. Moreover, through MetaNeigh-
combining AEs for each modality with a graph-based neural bor analysis between vertebrates and invertebrates, researchers
network, GLUE bridges the gaps among three different omics, constructed a cross-species cell-type evolutionary tree at the
including gene expression, chromatin accessibility, and DNA organism level for tracing the origin of major cell lineages during
methylation, thereby enabling the effective triple-omics integra- animal evolution (Wang et al., 2021b) and further identified
tion. common pathways that change across typical life span (Wang et
Another challenge in multimodal analysis is the need for al., 2023b) to shed light on the dynamic changes in cell function
integrating scRNA-seq data with ST data. The spatial informa- and regulation across species. However, MetaNeighbor analysis
tion of cells plays a crucial role in determining their states and has limited usage for remote species because they share a few
functions, which enables the reconstruction of transcriptional one-to-one ortholog genes. To address this concern, Tarashansky
panoramas for tissue architecture. Although ST datasets contain et al. (2021) proposed the self-assembling manifold mapping
transcriptome profiles along with spatial localization, they often (SAMap) algorithm that incorporates the relative contributions
lack high resolution for individual spots. On the other hand, of homologous genes derived from reciprocal BLAST for cross-
scRNA-seq data retains individual cell heterogeneity but species cell mapping. Using SAMap, researchers identified 1,209
eliminates the spatial information due to cell isolation proce- homologous cell-type pairs across eight species, and only 36.5%
dures. Therefore, there is an urgent need for new integrative (441 of 1,209) could be recognized by MetaNeighbor analyses
approaches to leverage complementary signals from both scRNA using one-to-one orthologous genes (Li et al., 2022c). In
and ST data. One application for integrative analysis of scRNA- addition, the result of benchmarking strategies for cross-species
seq and spatial expression data is in cellular devolution, which integration revealed the superiority of SAMap over BEC methods
helps uncover the cellular heterogeneity for each captured spot in in integrating whole-body or organism-level atlases among
ST data. Methods designed for cell devolution tend to employ species (Song et al., 2023). Notably, cross-species deep learning
corresponding scRNA-seq datasets as references, including CARD methods such as CAME (Liu et al., 2023c) and SATURN (Rosen

1252 SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 [Link]
et al., 2024) were constructed recently and demonstrated better epigenetic level, AI models are used to transfer labels from
performance than SAMap in terms of benchmarking. To deal scRNA-seq to scATAC-seq data, as demonstrated by scGCN (Song
with homologous genes, SATURN uses the concept of “macro- et al., 2021), and to construct TF regulatory networks from
genes,” defined as groups of genes with similar protein single-cell ATAC-seq data, as exemplified by DeepTFni (Li et al.,
embeddings, and CAME applies graph neural networks to make 2022b). At the spatial level, specially designed AI models are
full use of homologous information. increasingly available (Zeng et al., 2022). Methods including
However, these methods and all the efforts are largely limited DeepST (Xu et al., 2022a), STAGATE (Dong and Zhang, 2022),
to pairwise comparisons followed by integrative analyses for the and DeLTA2.0 (O’Connor et al., 2022) have been designed to
identification of homologous cell-type pairs across species. capture both gene expression profiles and spatial information.
Recently, Mah and Dunn (2024) employed principal components Furthermore, CellCharter (Varrone et al., 2024) has been
as phylogenetic features to reconstruct cell type evolution across developed to identify the cellular spatial niches related to tissue
species through cell phylogenies. Although cell phylogenies remodeling and cell plasticity. In addition, Graph Convolutional
allow researchers to phylogenetically define cell type via clade Neural Networks for Genes (GCNG) (Yuan and Bar-Joseph, 2020)
membership, they remain restricted to one-to-one homologies and HoloNet (Li et al., 2023a) are commonly used to infer cell-
shared by species. In addition, Wang et al. (2021b) inferred the cell interactions at a spatial resolution.
phylogenetic tree with IQ-TREE (Nguyen et al., 2015), indicating For a deeper exploration of gene regulatory programmers, it is
that cell types from the same cell lineage tend to be clustered in essential to construct AI models with million-level single-cell
vertebrate pairwise comparison, while, in invertebrates, cell types data. Big data-inspired AI models hold promise for unraveling
from the same species tend to cluster. Therefore, cross-species complex gene interactions and regulatory mechanisms, thereby
analysis utilizing traditional cell phylogenetic trees, especially for aiding in the understanding of disease processes and potential
homologous gene selection in lower species, remains challen- therapeutic targets. Recently, Geneformer (Theodoris et al.,
ging. Future work should consider the incorporation of 2023), a transformer-based model pretrained on a corpus of 30
paralogous and orthologous genes in constructing cell phylo- million scRNA-seq data, has been used to facilitate predictions in
genies to address this issue. settings with limited data using transfer learning. The pretrained
Overall, cross-species analyses enhance our understanding of Geneformer model demonstrated excellent performance in
the evolution of cell types. This analysis not only informs insights predicting gene dosage sensitivity. The fine-tuning procedure
into the genetic and functional relationships between different for downstream tasks includes predictions on chromatin
species but also paves the way for a better understanding of the dynamics and network dynamics. Specifically, in silico perturba-
fundamental principles underlying the diversity and adaptation tion includes in silico deletion and in silico activation. The ability
of cell types across various organisms. From the transcriptional to simulate these perturbations in silico is thus crucial as it can
similarity of homologous genes to the construction of cell provide a valuable reference point for guiding downstream
phylogenetic trees, cell type evolution is increasingly getting experimental work, facilitating more efficient and effective
integrated into the evolutionary perspective. biological research. Similarly, the AI model scGPT (Cui et al.,
2024) is pretrained on over 10 million cells and enables multi-
omics integration. Additionally, the transformer-based cross-
AI application for single-cell genomics
species foundation model GeneCompass (Yang et al., 2023b) is
Single-cell technologies generate large amounts of data, typically pretrained on an extensive scRNA-seq dataset of more than 120
with characteristics of high dimensionality, sparsity, and million cells from humans and mice. Notably, single-cell
complexity, which makes the analysis using conventional transformers models have exhibited versatility in single-cell
computational methods difficult and unfeasible. To address these genomics, as evidenced by the resources available on the web:
challenges, researchers are turning to AI methods as potential [Link]
alternatives. (scBERT) (Yang et al., 2022). These models can tackle a wide
The application of AI algorithms in the field of single-cell range of downstream tasks, including cell type annotation, gene
genomics is expected to offer significant advantages for analysis, expression imputation, genetic perturbation, cell clustering, gene
including batch effect correction, denoising, and dimensionality regulatory network inference, simulation, and denoising. Cur-
reduction, making the single-cell genomics datasets easier to rent single-cell transformer models can be categorized into
interpret and visualize. AI algorithms method can mainly use various groups. In conventional transformer architectures,
two types of data as input: non-sequential data (such as cell models like scBERT, BioFormers (Fang et al., 2023), and CellLM
profiles and gene expression from scRNA-seq) and sequential (Zhao et al., 2023) utilize standard transformer architectures,
data (such as DNA sequences of cCREs from scATAC-seq). often employing Masked Language Modeling (MLM) with
Currently, AI algorithms can holistically account for all stages different loss functions and objectives for gene expression
of single-cell genomics analysis (Brendel et al., 2022; Erfanian et prediction or cell representation learning. Decoder-based models
al., 2023). For example, the applications of the AI models such as scMulan (Bian et al., 2024) and tGPT (Shen et al., 2023)
mentioned earlier include several categories: deconvolution rely on decoder architectures for tasks such as conditional cell
(such as TAPE (Chen et al., 2022c)), BEC (such as SAUCIE generation or batch integration. Multiencoder models such as
(Amodio et al., 2019) and scGen (Lotfollahi et al., 2019)), GeneCompass and scFoundation (Hao et al., 2024a) utilize
integration of multimodal omics (such as GLUE (Cao and Gao, multiple encoders, often to handle different aspects of the data,
2022)), and cross-species analysis (such as CAME (Liu et al., such as genetic perturbation effect prediction. In AE models, the
2023c)). Other applications include cell-cell communication multitask learning model (Pang and Tegnér, 2020) employs an
analysis (such as DeepCCI (Yang et al., 2023a)) and RNA AE with two transformer encoders for cell clustering. The
velocity (such as DeepVelo (Chen et al., 2022e)). At the Universal Cell Embedding (Rosen et al., 2023) incorporates

[Link] SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 1253
specialized embeddings, which are designed to capture informa- atlases across numerous species. Typically, a conventional cell
tion about the gene expressions or chromosomal positions, atlas should, at a minimum, depict a comprehensive catalog of
enhancing their ability to learn meaningful representations of cell types in either single or multiple tissues across at least one
cells. iSEEEK (Shen et al., 2022) and Geneformer leverage rank- dimension (i.e., transcriptome, genome, proteome, or metabo-
based ordering of gene expression values within their input lome). Considering the importance of the transcriptome, the
representations, thereby contributing to their performance in majority of current cell atlases are constructed from the
gene-related tasks. scFormer (Li et al., 2024) is a great example of transcriptomic level. As a preeminent animal model, high-
a task-specific model that is designed to address specific throughput single-cell RNA sequencing (scRNA-seq) methodol-
challenges in single-cell analysis by using a modified MLM ogies have been primarily utilized for the pioneer mapping of
approach coupled with other techniques. Models such as TOSICA murine cellular atlas (Mus musculus). The cellular compositions
(Chen et al., 2023b), scTransSort (Jiao et al., 2023), STGRNS (Xu of diverse murine tissues serve as a reference for studies
et al., 2023c), CIForm (Xu et al., 2023b), and TransCluster (Song investigating developmental, aging, and disease mechanisms.
et al., 2022) represent a diverse category of unsupervised Despite the efforts to characterize the single-cell landscapes
approaches, each with its own set of objectives and unique across various tissues (Haber et al., 2017), it was not until 2018
embedding strategies. Overall, big data-inspired AI models that the MCA unveiled an exhaustive single-cell transcriptomic
implement comprehensive analyses across various biological profile encompassing nearly half a million cells across more than
applications. 40 organs and tissues (Han et al., 2018b). This repository stands
However, the abovementioned AI models operate on single-cell as a pivotal resource conducive to the discernment of cell types
data independent of the knowledge about the sequence, such as across predominantly studied organs as well as unexplored
the gene expression profiles. Sequence-based AI models have tissues. The Tabula Muris Consortium reported another compre-
emerged recently, offering direct modeling of the genome hensive single-cell transcriptome atlas of 20 mouse organs
sequentially. These models offer the potential to analyze and (Tabula Muris et al., 2018). The use of a combination of high-
understand genetic information in a more comprehensive and sensitivity full-length alongside 3′-end methods has facilitated a
detailed manner, thereby providing valuable insights for various more detailed characterization of TF patterns across diverse cell
applications in genomic studies and personalized therapies. Nvwa lineages. These studies serve as a pioneering landmark, defining
(Li et al., 2022c) is a groundbreaking CNN model using 13-kb the first large-scale cell atlas spanning multiple murine tissues at
DNA sequences as inputs. It creates a unified framework to the gene level. In the realm of chromatin accessibility,
analyze the regulatory process of gene expressions and elucidate Cusanovich et al. (2018) pioneered the first single-cell chromatin
the mechanism of how different cell types use the same DNA accessibility atlas on a genome-wide scale, encompassing 13
sequence to encode different genes. Nvwa effectively utilized AI- distinct adult murine tissues. In addition, the presence of
derived cis-regulatory elements as well as identified complex regulatory elements, being cell type-specific, provides a robust
regulatory rules, presenting a novel strategy for studying framework purposed for pinpointing cell types that are pre-
regulatory grammar in diverse biological systems. Another dominantly associated with common human diseases and traits.
model based on the CNN architecture, scBasset (Yuan and Other representative studies focused on cell heterogeneity in
Kelley, 2022), inputs 1,344-bp DNA sequences from scATAC-seq different systems. For example, Fang et al. (2022b) integrated the
data. scBasset achieves cell clustering, data denoising, data stromal cell atlas across multiple tissues during the develop-
integration across assays, and TF activity inference. In particular, mental stages. Zeisel et al. (2018) mapped the single-cell
scBasset can predict perturbations including single nucleotide transcriptome atlas of the murine mammalian nervous system
mutations as well as TF motif insertions in the non-coding in major anatomical regions, including the brain, spinal cord,
regions. To bridge the gap between epigenetic regulation and peripheral sensory organs, and the enteric and sympathetic
gene expression, AI models such as Huatuo (Xiao et al., 2023), nervous systems. In addition, Kalucka et al. (2020) constructed
ExpectoSC (Sokolova et al., 2023), and seq2cell (Schwessinger et the first mouse endothelial cell atlas and established metabolic
al., 2023) are constructed by integrating sequence-based signatures in different vascular beds.
epigenetic models to facilitate the prediction of cell type-specific The murine embryo serves as a homologous model that is
expression directly from DNA sequences. Huatuo and ExpectoSC instrumental in unraveling the integral molecular events
are constructed on the framework of CNN-based DeepSEA Beluga entwined with the process of cell development. After the release
(Zhou et al., 2018), while transformer-based Enformer (Avsec et of the adult mouse cell atlas, another two studies constructed a
al., 2021) is used in seq2cell. Notably, these AI models promote single-cell transcriptome atlas in 2019, with a focus on mouse
the estimation of the effects of single nucleotide mutations on the gastrulation and organogenesis. Pijuan-Sala et al. (2019)
gene expressions. However, Sasse et al. (2023) proposed that adopted 10x genomics technology to comprehensively delineate
current sequence-based AI models often fail to accurately predict the transcriptomic landscape across nine distinct temporal
the direction of mutation effects on the gene expression. These points, spanning from 6.5 to 8.5 days postfertilization. In a
authors suggest leveraging of diverse genomes and biochemical complementary study, Cao et al. (2019) innovatively employed
processes as inputs for future deep learning models. sci-RNA-seq3 technology to analyze approximately 2 million
cells derived from 61 embryos, spanning stages from 9.5 to 13.5
days of gestation. Collectively, these two studies provided
Chapter 6. Advances in single-cell atlas mapping
continuous and complete lineage-specific gene expression
resources. As for another mouse developmental cell atlas, Fei et
Mouse cell atlas
al. (2022) unveiled a comprehensive mouse cellular differentia-
The evolution and enhancement of throughput and sensitivity in tion atlas that captures the temporal series of morphogenesis,
single-cell technologies have facilitated the development of cell extending from the early embryogenesis stage at E10.5 to

1254 SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 [Link]
postnatal day 21. The Xbp1 was identified as a pivotal master atlases of the fetal-maternal interface, thereby adding another
regulator governing cell fate decisions in mice. Mittnenzweig et feather of accomplishment to their ongoing scientific exploits
al. (2021) reported a single-embryo, single-cell-level mouse (Greenbaum et al., 2023; Hickey et al., 2023; Lake et al., 2023).
gastrulation (E6.5 to E8.1) atlas. Qiu et al. (2024) constructed Research communities in China are keen participants who are
the largest scRNA-seq dataset to date of mouse embryonic making salient contributions toward shaping the trajectory of
development and presented a comprehensive atlas (Ontogeny of HCA research. In an epoch-making study published in 2020,
Mouse) consisting of an impressive 12.4 million cells, covering Han et al. (2020) presented the world’s foremost single-cell
critical stages spanning from gastrulation to the postnatal period. transcriptome landscape that spans multiple tissues—a colossal
Other single-cell multi-omics studies have supplemented results dataset comprising over 700,000 cells from an excess of 50 tissue
from other dimensions, including cis-elements, histone modifica- types—thereby providing an insight into the comprehensive
tions, and spatial information (Gorkin et al., 2020; Jiang et al., organ system of the human body. This research unprecedentedly
2023b; Smith et al., 2022). In the dimension of STs, Peng et al. encompasses data about a comprehensive range of adult tissues,
profiled the spatial transcription atlas of early gastrulation embryonic structures, as well as diverse cell lines. In parallel,
(embryonic day 5.5 (E5.5) and E7.5), as well as the brain they unfolded a thorough single-cell transcriptional profiling
development (Jiang et al., 2023a; Peng et al., 2019a). Other study covering an expanse of 15 organs derived from a standard
researchers adopted Slide-seq and Stereo-seq to portray mouse adult individual (He et al., 2020). Progressing along the
whole embryo organogenesis at the subcellular resolution, chronological thread of research within the sphere of multi-
respectively (Chen et al., 2022a; Sampath Kumar et al., 2023). organ cell atlas, the Tabula Sapiens Consortium has contributed
Altogether, these two studies covered the development stages significantly in the scientific realm, crafted another atlas that
from E8.5 to E16.5. The open-access ST databases greatly encompasses multiple organs at a single-cell transcriptomic level,
facilitated other researchers in exploring these resources. Some and distinctly identified nearly 500 cell types (The Tabula
reports have already been published in the field of aging research. Sapiens Consortium et al., 2022). Furthermore, Eraslan et al.
For instance, the “Tabula Muris Senis” highlights aging-related (2022) employed single-nucleus RNA sequencing as an effica-
perturbations and systemic inflammation in multiple tissues cious stratagem to scrutinize frozen, banked specimens pertain-
(Almanzar et al., 2020). The spatial and single-cell atlas of the ing to eight human organ types procured from 16 philanthropic
cycling and aging female reproductive tract sheds light on the donors. This sophisticated exploration proposed an innovative
role of fibroblasts and fibrosis in extracellular matrix reorganiza- framework for the utilization of frozen tissue samples, thereby
tion during a reproductive lifespan (Winkler et al., 2024). significantly augmenting the Genotype-Tissue Expression (GTEx)
project’s repertoire.
The Human Developmental Cell Atlas (HDCA), a significant
Human Cell Atlas (HCA)
component of the HCA, is launched with the distinct objective of
Initiated in 2016, the HCA consortium signifies a worldwide formulating an exhaustive reference map at the single-cell level
collaborative endeavor, diligently working to characterize a to capture the nuances of normal organogenesis and early
cellular reference blueprint of all human tissues (Regev et al., developmental stages. This ambitious undertaking has laid a
2018). Enlisting the utilization of avant-garde single-cell robust foundation to further the understanding of aging and
genomic methodologies, this collective effort aims to profoundly cancer and to invigorate the field of regenerative medicine
understand human health and various diseases to provide (Haniffa et al., 2021). Within this realm, the laboratories of Qiao
insights into the fundamental mechanisms shaping our physio- and Tang have conducted a series of pioneer multi-omics studies,
logical and pathological states. The first draft of the HCA aims to charting new territories in understanding human germline cells
profile over 100 million human cells from all organs in men and and embryo implantation processes (Li et al., 2017; Li et al.,
women. In this draft, the data resource will mainly be based on 2018; Zhou et al., 2019). With the advancement and
the mature scRNA-seq (also including other omics) profiling of popularization of scATAC-seq methods, an increasing number
cell suspension (Lindeboom et al., 2021). The end version will be of HCA studies are concurrently focusing on the cell type-specific
a comprehensive, super-resolution spatiotemporal multiscale regulatory elements brought about by chromatin accessibility
omics database encapsulating the dynamic landscape of every data. Past research has followed along various stages of human
single cell type within the human body. This seminal resource, embryonic development. Two such in-depth studies have
which is readily accessible to the scientific community, promises produced comprehensive transcriptome and spatial datasets that
to serve as a reference atlas of inestimable value not only for basic cover critical phases in human embryogenesis. These phases
biological discoveries but also as an instrumental tool for an encompass the early gastrulation period, specifically 16- and 19-
array of clinical applications. The development of the HCA days postfertilization, as well as the organogenesis stage,
community is also expected to facilitate the translation of occurring 4 to 6 weeks into fetal development. These studies
technology from basic research to clinical application scenarios. have provided an intricate understanding of these crucial
In 2019, marking another significant milestone, the National development stages (Tyser et al., 2021; Xu et al., 2023d). The
Institutes of Health (NIH) instituted the Human Biomolecular group’s Genomic Architecture of Cells in Tissues (GeACT)
Atlas Program (HuBMAP). This endeavor inaugurated an project, alongside a comparable preprint multi-omics study,
additional research consortium dedicated to charting the terrain employed high-detectability scRNA-seq and scATAC-seq ap-
of the human body at the single-cell resolution (Snyder et al., proaches. They systematically profiled the cellular atlas of
2019). In the year 2023, the HuBMAP project rendered a report human fetuses during gestation, focusing on the window of 8–
encapsulating its advancements, as signified by the depiction of 21 weeks (Tian et al., 2020; Yu et al., 2021a). Shendure’s
single-cell transcriptomic atlases for the human kidney and laboratory pioneered the creation of the most extensive fetal HCA
intestines. Moreover, it explored new frontiers by providing ST to date (transcriptome and chromatin accessibility). This

[Link] SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 1255
groundbreaking work catalogs 4 million individual cells derived atlas that intricately mapped the single-nucleus transcriptomic
from a spectrum of 28 fetuses, with estimated postconceptual age landscape of the entire macaque cortex (Chen et al., 2023a). A
ranging from 72 to 129 days (Cao et al., 2020a; Domcke et al., comprehensive cross-species transcriptomic data comparison
2020). Zhang et al. (2021b) pioneered the inaugural scATAC- comprising humans, macaques, and mice underscored the
seq atlas, elucidating over one million nuclei from a diverse set of presence of cell types exclusive to primates, predominantly
222 distinct cell types across 30 different adult human tissues. enriched within the fourth layer of cortical tissues.
These authors also performed an integrative analysis of previous Zhai et al. (2022) provided a comprehensive report on the
published fetal datasets and their study provided a research gastrulation and early organogenesis cellular atlas of monkeys,
paradigm to identify the correlation of fetal and adult human which notably bolstered our comprehension of primate develop-
cCREs with human traits and diseases. With the advancements of ment. Their study demonstrated the transcriptomic character-
ST technology in recent years, these methodologies have now istics pertinent to diverse cell types during primitive streak
been incorporated into the mapping of HCAs to validate key expansion, somitogenesis, the genesis of the gut tube, neural tube
molecular feature events during developmental processes with patterning, and neural crest differentiation periods. Building
temporal and spatial resolution. Pan et al. (2023), for instance, upon this body of research, another study combined scRNA-seq
reported the inaugural spatiotemporal transcriptome atlas of and scATAC-seq to construct a detailed map of gene expression
human embryos post-gastrulation. This pioneering research and chromatin accessibility at the single-cell level across 16
effort revealed organ-centric regulons that are postulated as organs in cynomolgus monkeys (Qu et al., 2022). Cross-species
potential determinants of cellular lineage. Moreover, they comparative investigations, encompassing mice, monkeys, and
succeeded in categorizing the genes of specific cell types humans, have proposed a distinctly elevated equivalence in
associated with developmental irregularities and infections, immune-related cellular characteristics, which is predominantly
thereby contributing significantly to our current state of knowl- shared between monkeys and humans. In another critical brain
edge about embryonic development and disease pathology. region, the adult and aging cellular atlases of the hippocampus
The advancements in single-cell atlases of various human have also been reported (Hao et al., 2022; Zhang et al., 2021a).
organ tissues are indeed remarkable and captivating. Recently, The authors unearthed the onset of proinflammatory reactions in
Jiao’s group (Li et al., 2023e) combined scRNA-seq and Stereo- aged microglial and oligodendrocyte cells to cultivate an adverse
seq to build the spatiotemporal developmental atlas of multiple microenvironment impeding neurogenesis. Moreover, the group
human brain regions from 6–23 gestational weeks. This is the led by Liu has made notable strides in recent years by unveiling a
most extensive spatiotemporal developmental transcriptome remarkable series of primate tissue aging cell atlases. They
atlas of the human brain to date, covering the greatest period assembled and established a unique single-cell transcriptome
and area; it uncovered the instrumental roles of diverse atlas of numerous aging primate organs, inclusive of the
interactions between different types of glial cells and neurons in cardiopulmonary system, ovaries, liver, testes, muscles, arteries,
the regional specialization process of neuron subtypes. Notably, and the pancreas (Huang et al., 2023a; Jing et al., 2023; Li et al.,
the Brain Research through Advancing Innovative Neurotech- 2021a; Ma et al., 2021; Wang et al., 2020b; Yang et al., 2024;
nology’s Initiative-Cell Census Network (BICCN) alliance, pro- Zhang et al., 2020c). Most aging-associated transcriptional
pelled by the NIH, has achieved a series of progressions in the alterations, which have been identified specific to certain cell
exploration of cell types and functions in humans, non-human types, underscored an amplified systemic inflammation as a
primates, and rodent brains. This project revolves around the characteristic signature of primate aging. Moreover, this group
construction and comparative analyses of single-cell atlases for ingeniously formulated tissue-specific therapeutic strategies
these species’ brain maps. These multi-omics works cover the geared toward intervening in the degenerative processes related
comprehensive brain cell types in different anatomic regions to aging.
(Braun et al., 2023; Chartrand et al., 2023; Johansen et al., Within the domains of piscine and amphibious studies,
2023; Jorstad et al., 2023a; Jorstad et al., 2023b; Kim et al., organisms such as zebrafish, Xenopus, and axolotl serve as
2023a; Lee et al., 2023; Li et al., 2023f; Siletti et al., 2022; Tian quintessential models, offering valuable insights into the intricate
et al., 2023b; Velmeshev et al., 2023; Weninger and Arlotta, processes of embryonic development and holistic tissue regen-
2023). Collectively, these efforts represent significant contribu- eration. Jiang et al. (2021) unveiled a comprehensive zebrafish
tions to further our understanding of human biology and pave cellular landscape, representing over 250,000 unique cells that
the way for future research in areas such as disease mechanisms, span across both embryonic and adult life stages. The authors
diagnostics, and therapeutics (Figure 8). shed light on the distinctive characteristics of the blastema
population, which are a crucial component of the zebrafish’s
extraordinary caudal fin regeneration. Briggs et al. (2018)
Cell atlas of other organisms
derived a comparative cell atlas of Xenopus and zebrafish
The nonhuman primate serves as another important model development to identify embryonic cell states. The first single-
organism in the field of biomedical research. It has proven to be cell transcriptome atlas of Xenopus, including multiple larval and
particularly instrumental in investigating human physiology, adult organs, was reported by Liao et al. (2022). In their study,
disease progression, and, notably, the intricacies of organ aging. the authors discovered a ubiquitous regulatory mechanism
The BGI in China generated the first nonhuman primate cell atlas encompassing the phenotype of antigen processing and pre-
(NHPCA) (Han et al., 2022). In their study, they provided an sentation, which were consistently manifested in cells derived
expansive cell transcriptomic atlas that incorporated data from from disparate embryonic layers during metamorphosis. The
more than 1 million cells spanning 45 tissues collected from Axolotl serves as another paradigm in the realm of develop-
adult nonhuman primates, specifically Macaca fascicularis. Sig- mental and regenerative biology. Notable single-cell RNA-
nificantly, the authors further established a high-definition ST sequencing studies investigating Axolotl limb regeneration

1256 SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 [Link]
Figure 8. Mapping of human and mouse cell atlases at the single-cell level.

uncovered distinctive subsets of stem cells and identified pivotal stage, the developmental cell atlas of C. elegans incorporated
regulatory activity within fibroblasts linked to the instigation of extensive whole-body cellular representations encapsulated
regeneration (Gerber et al., 2018; Lin et al., 2021b). Further- within a singular organism (Cao et al., 2017). Meanwhile, the
more, the inaugural atlas of axolotl cells provided a comprehen- lineage-resolved molecular atlas of C. elegans during embryogen-
sive mapping of transcriptomic alterations that occurred esis and subsequent larval development evidenced a pattern of
throughout the course of metamorphosis and development (Ye lineage priming within distinct cells right up unto the precipice of
et al., 2022). In the most recent advancements, an ST dataset terminal fate determination (Packer et al., 2019). In addition,
indicated an injury-induced cluster of ependymal cells that some review articles have encapsulated the application and
manifested as a progenitor cell population with the ability to advancements of single-cell technologies within the realm of
replace lost neurons (Wei et al., 2022). This research further plant studies (Bawa et al., 2022; Yu et al., 2023b).
delineated the presence of a state resembling immature neurons
at the injury site.
Chapter 7. Single-cell insights for translational
The fruit fly (Drosophila melanogaster) is a classic model
medicine
organism in the study of neurobiology and genetics. Li et al. (Li
et al., 2022a; Lu et al., 2023) diligently constructed a cross-tissue Single-cell genomics has emerged as a powerful tool in decipher-
single-cell transcriptome atlas encompassing both adult and ing the complexities of human diseases, particularly in the field of
aging flies that featured the gene signatures and correlating TFs cancer. By providing unprecedented resolution and insights into
manifesting across various cell types. Serving as a point of cellular heterogeneity, single-cell genomics has revolutionized
reference, these cellular signatures catalyze further explorations our current understanding of disease mechanisms, treatment
into genetic perturbations and disease modeling studies. response, and patient stratification, as well as drug discovery and
Furthermore, the aging fly cell atlas sheds light on the evolution therapy selection in the context of precision medicine. In this
of age-related gene expression and the dynamic changes in section, we have highlighted how single-cell genomics can be
pathways, such as the noticeable shift in lineage toward the fat translated into clinical applications and discussed its transforma-
body and the incidence of apoptosis in muscles. In a noteworthy tive impact on disease understanding and treatment strategies
addition, the authors established an aging clock model that (Figure 9).
ingeniously allows the prediction of an animal’s age purely based
on scRNA-seq data. Fly brain cell atlas has successfully
Transformative impact of single-cell genomics on cancer
deciphered the single-cell accessible chromatin and transcrip-
treatment
tome atlas throughout adulthood and during the aging process,
thereby identifying comprehensive regulatory regions associated Hyperfine resolution of single-cell genomic: better targeting cancer
with the processes of developmental trajectories, neurogenesis, cells in tumor progression
reprogramming, and maturation (Davie et al., 2018; Janssens et As a leading cause of death across the world, cancer poses
al., 2022b). The comprehensive whole-body cell landscapes formidable dangers with a widespread menace in mechanisms
encompassing zebrafish, flies, and earthworms have been that are yet to be disentangled. Dissecting cancer mechanisms is
painstakingly assembled thereafter, paving the way for a of paramount importance in understanding the complexities of
genomic sequence-grounded, cross-species deep learning model. cancer biology, with pivotal implications in diagnosis, treatment
This state-of-the-art model holds the potential to accurately strategies, and prognostic predictions. Thanks to the unprece-
predict cellular lineages for creating new avenues for future dented aid of single-cell genomics, finer profiling of cancer cells
research (Li et al., 2022c). Certain cell atlas mappings of provides deeper insights into the treatment and mechanisms of
terrestrial invertebrates have been recorded in the existing cancer, which encompass several key aspects such as cancer
scientific publications. The cell atlas of the earthworm manifested initiation and progression, cancer metastasis, and cancer
a more comprehensive array of major cellular lineages than heterogeneity.
those examined previously (Shao et al., 2020). At the L2 larval Cancer initiation and progression are complicated processes

[Link] SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 1257
Figure 9. Clinical applications of single-cell genomics in cancer treatment (A), treatment of non-cancer diseases (B), and drug discovery and precision medicine (C).

that involve the acquisition of genetic and epigenetic alterations specific mutations and epigenetic changes that confer a selective
by normal cells resulting in the development of malignant advantage, resulting in the initiation and propagation of cancer
properties and, finally, modifying the surrounding host cells into (Frank et al., 2021; Lawson et al., 2015; Yang et al., 2017).
the tumor-supportive tumor microenvironment (TME). Cancer Single-cell omics technology has the potential to unveil the
driver genes, whose mutations are more likely to induce cancer intricate cellular and molecular dynamics of cancer cells under-
than those of passenger genes (Greenman et al., 2007), and lying cancer metastasis, which refers to the spread of cancer cells
cancer stem cells (CSCs), which arise by mutation from normal from the primary tumor site to distant organs or tissues—a
stem cells and maintain the self-renewal capability (Jordan et al., complex process encompassing multiple biological events invol-
2006), are thought to be two key initiators of tumorigenesis ving invasion, intravasation, dissemination, extravasation, and
(Hanahan, 2022). Single-cell DNA sequencing (scDNA-seq) and colonization (de Visser and Joyce, 2023; Gerstberger et al.,
its mapping to scRNA-seq can identify somatic mutations, copy 2023). By employing single-cell spatial, transcriptomic, and
number alterations, and chromosomal rearrangements in epigenomic analyses, several recent studies can successfully
individual tumor cells, thereby revealing the phylogenetic elucidate the immune landscape and intricate cellular interaction
relationships and temporal order of mutational events (Funnell network within metastatic tumors (Karimi et al., 2023; Zhang et
et al., 2022). Single-cell omics technologies also illuminate the al., 2023b). Furthermore, CTCs and disseminated tumor cells
isolation and identification of CSCs, which helps understand the (DTCs) represent tumor cells that respectively detach from the

1258 SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 [Link]
primary tumor and enter the bloodstream or lymphatic system the cellular components and diversity of TME, some of which
(Lin et al., 2021a) as well as those that have already shed from possess antitumor characteristics and others play pro-tumor
the primary tumor and migrated to other tissues or organs (Kang roles.
and Pantel, 2013). Understanding the mechanisms of how CTCs Dissecting the anti-tumor response of various T cell types in the
and DTCs survive physical shear stress, evade anoikis, and evade context of TME is the key to enhancing the efficacy of cancer
clearance by immune cells is crucial in the field of metastasis treatment, which, therefore, entails rigorous and comprehensive
research (Gerstberger et al., 2023; Ring et al., 2023). Single-cell profiling across multiple dimensions. When combined with TCR
genomics facilitates the characterization of CTCs at the cellular sequencing, scRNA-seq empowers the identification of various T
level in the context of their rare nature and uncovering their cell subtypes in pan-cancer landscapes (Zheng et al., 2021). For
heterogeneity, plasticity, survival strategies, and metastatic example, exhausted CD8+ T cells (Tex), featured with reduced
capacity (Liu et al., 2023d; Reza et al., 2021; Suvilesh et al., cytotoxicity and effector function (Blank et al., 2019; Chen and
2022; Ye et al., 2023). In addition, the employment of single-cell Flies, 2013; Wherry, 2011), have been identified to demonstrate
sequencing technology for the detection of CTCs has made a preferential enrichment in various tumors (Oliveira et al., 2021;
noteworthy impact on facilitating early-stage diagnosis and Zhang et al., 2018; Zheng et al., 2017a), with LAYN as a
prognostic monitoring of oncology. signature gene exerting negatively regulation functions. Mean-
Tumor heterogeneity, the diversity of cancer cells within a while, the CXCL13+ activated precursors of exhausted CD8+ T
tumor or among different tumors, serves as a pivotal determinant cells have been associated with tumor-reactive features and a
in tumor evolution, treatment response, and drug resistance and better prognosis (Guo et al., 2018; Liu et al., 2022b), while
can arise from genetic variations, epigenetic modifications, or CXCL13+CD4+ T cells, also denoted as T helper 1 (Th1) like T
TME factors (Prasetyanti and Medema, 2017). Deciphering cells, seem to have anti-tumor potentials (Zhang et al., 2021d;
tumor heterogeneity is paramount for the selection of targeted Zheng et al., 2022), both of which collectively play a predictive
therapies determined based on molecular features. In early-stage role in immune checkpoint blockade (ICB) response (Liu et al.,
investigations, tumor heterogeneity was originally defined based 2022b). Another subtype of CD4+ T cells, regulatory T cells
on discrepancies in driver gene mutations among distinct tumor (Tregs), have been characterized into two clusters in single-cell
types, reflecting inconsistent mutation patterns and frequencies profiling of nonsmall cell lung cancer (NCSLC) based on the
(Wood et al., 2007). This concept gradually evolved to expression pattern of TNFRSF9 (4-1BB), whose TNFRSF9+
encompass the heterogeneity of mutation profiles within the cluster highly express genes such as IL1R, which is associated
same tumor, implying intratumoral diversity of neoplastic cells with immunosuppressive functions (Guo et al., 2018).
(Gerlinger et al., 2012), such as the absence of EGFR mutations Myeloid cells, which are comprised of several distinct popula-
in some lung cancer cells, which limits the efficacy of EGFR- tions, account for substantial proportions and mass of immune
targeted therapies in eradicating most cancer cell population. cells and are acknowledged as comparable contributors to tumor
Subsequently, with the advancements in diverse single-cell progression and metastasis (Sender et al., 2023), thereby
technologies, the integration and temporal analysis of numerous attracting continuous scrutiny as potential targets for cancer
single-cell pan-cancer atlases have given a more nuanced therapeutic interventions (Engblom et al., 2016). However, it is
definition of tumor heterogeneity encompassing the comprehen- imperative to note that not all subtypes of myeloid cells within
sive illustration of differential composition and dynamic interplay the TME exhibit pro-tumor function, thereby necessitating a
in the TME across distinct tumor types, disparate tumor regions, pressing demand for the application of single-cell genomics to
and among different patients (Barkley et al., 2022; Cheng et al., dissect their heterogeneity and delineate the immunosuppressive
2021b; Tang et al., 2023a; Zheng et al., 2021). Moreover, single- clusters for more precise therapeutic targets (Cheng et al.,
cell multi-omics technologies, such as scATAC-seq, scTrio-seq, 2021b). Multiple discrete single-cell studies have autonomously
and CITE-seq, provide supplementary perspectives for disentan- identified several pivotal subtypes of myeloid cells, such as C1QC+
gling the intricacies of tumor heterogeneity (Bian et al., 2018; and SPP1+ macrophages, which exhibit inflammatory and
Satpathy et al., 2019). Hence, beyond the mere characterization angiogenic signatures, respectively (Zhang et al., 2020a).
of cancer cell genetic features, unraveling the heterogeneity and Notably, neither of these subtypes aligns with the conventional
intricate interactions within the TME via single-cell genomics M1 and M2 dichotomous phenotypes, thus underscoring the
assumes a critical and indispensable role. indispensability of employing single-cell technologies for precise
macrophage classification. In hepatocellular carcinoma (HCC),
Unlocking new frontiers through single-cell genomics: the potential of single-cell analysis has contributed to the delineation of specific
targeting tumor-specific TME cell subtypes subgroups within tumor-associated neutrophils (TANs), such as
Our understanding of cancer has evolved over the course of time, CCL4+, SPP1+, and PD-L1+ TANs, which have demonstrated
shifting from a narrow focus concentrated on cancer cells to a pro-tumor attributes and potential as therapeutic targets (Xue et
broader understanding encompassing the intricate surrounding al., 2022). Also designated as mregDC, LAMP3+ DC has been
ecosystem composed of non-malignant cells and the extracellular shown to possess the potential to migrate to lymph nodes and
matrix, termed the TME. Despite their profound impact on the regulate multiple T cells and NK cells through crosstalk (Maier et
discovery of cancer driver genes, traditional technologies for bulk al., 2020; Zhang et al., 2019c), such as by regulating the ICB
samples, which treat the tumor sample as a cohesive unit and response through cellular triads formed by mregDC, CXCL13+
primarily capture reads from cancer cells, fail to portray the Th, and PD-1hi Progenitor CD8+ T cells (Magen et al., 2023),
detailed components and tumor-associated signals of the holistic thereby potentially regulating the anti-tumor functions of NK
tumor ecosystem. Thanks to their capability of deciphering cells (Tang et al., 2023a).
cancer genomes at a hyperfine resolution, burgeoning single-cell Immune and stromal cells other than T cells and myeloid cells
technology has elevated the cancer research paradigm to reveal also hold paramount significance in the process of cancer

[Link] SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 1259
procession. A recent pan-cancer research presented a panorama and endothelial subpopulations linked to ICB resistance, there-
of human natural killer cells by analyzing their heterogeneity by opening up new avenues for immunotherapeutic exploration
and function in tumors from 716 patients with 24 cancer types, (Kang et al., 2022). Among the immune constituents within the
uncovering a subset denoted TaNK cells, which are enriched in TME, myeloid cells dominate, with specific subsets, including
tumors and exhibit anti-tumor function and unfavorable tumor-associated macrophages (TAMs), exhibiting suppressive
prognosis (Tang et al., 2023a). Another research demonstrated properties that contribute to ICB resistance. In an in-depth
that LRRC15+ cancer-associated fibroblasts (CAFs), as influenced scrutiny of HCC, single-cell data uncovered PD-L1+ TANs as a
by TGF-β, promote tumor growth and exert a direct inhibitory potential immunotherapeutic target, considering their capacity
effect on CD8+ T cell function to restrict the ability to respond to to impede T cell cytotoxicity (Xue et al., 2022). A multi-
checkpoint blockade therapies (Krishnamurty et al., 2022). dimensional single-cell analysis revealed the crucial involve-
ment of SPP1+ TAMs through the SPP1-CD44 axis and the
Uncovering immunotherapy biomarkers and prospectively enhancing targeting potential of enhancing the efficacy of immunotherapy
response rates with single-cell genomics in driving tumor immune evasion in AFP-positive HCC (APHC)
ICB therapies, such as anti-PD-1/L1 treatment, have demon- (He et al., 2023).
strated significant efficacy in controlling tumor progression in
some cancer patients (Sharma et al., 2023). However, the Optimizing therapeutic strategies for chemotherapy, radiotherapy,
response rates vary among different tumor types, and a and combination therapies
considerable proportion of patients do not benefit from these The mechanisms underlying the efficacy of chemotherapy and
treatments (Nishino et al., 2017). Moreover, some of the radiotherapy in clinical cancer treatment have extended beyond
currently identified biomarkers, such as tumor mutational the cytotoxic effects to now include the modulation of the anti-
burden, mismatch repair (MMR) deficiency, and PD-L1 expres- tumor immune microenvironment (Galluzzi et al., 2015).
sion, do not consistently predict treatment outcomes across all Leveraging single-cell genomics methodologies has facilitated
cancer types. Consequently, there is a pressing need for the the exploration of latent mechanisms, thereby advancing the
establishment of precise biomarkers that can accurately forecast formulation of enhanced therapeutic strategies for chemother-
the therapeutic response to ICB to enable optimal patient apy, radiotherapy, and combination treatments, thereby poten-
selection and maximize efficacy with minimal toxicity. By tially optimizing the anti-tumor outcomes. In gemcitabine plus
enabling high-resolution analyses of dynamic changes in the cisplatin treatment of nasopharyngeal carcinoma, for instance,
TME before and after treatment, single-cell sequencing technol- single-cell analysis showed that chemotherapy induces B cells to
ogy can facilitate the exploration of immunotherapy mechanisms stimulate the expansion of TFH and TH1 through IL-1β.
and help identify novel and effective biomarkers. Notably, a past Remarkably, IL-1β has been identified as a potential biomarker
study investigating PD-1 antibody treatment in basal cell for exploring the treatment efficacy, promisingly enhancement of
carcinoma revealed the concept of clonal replacement, wherein patient management, and informing the selection of therapeutic
the therapy improved the numbers of tumor-specific CD8 T cells approaches (Lv et al., 2023). An exploration of chemotherapy
with novel clonotypes (Yost et al., 2019). Another study in data in breast cancer confirmed the post-chemotherapy enrich-
NSCLC through scRNA-seq and scTCR-seq analysis identified an ment of ICOSL+ B cells, a subset that enhances chemotherapy
expansion of precursor-exhausted T cells among treatment efficacy by eliciting an anti-tumor T cell immune response with
responders, originating from local expansion or peripheral blood its functionality intricately regulated by the complement-CR2
T cell infiltration and replenishment, which was defined as clonal signal (Lu et al., 2020). This past study also posits CD55 as a
revival, expanding upon the theory of clonal replacement (Liu et potential target for enhancing chemotherapy effectiveness.
al., 2022a). Subsequently, through a meta-analysis of single-cell Single-cell analysis has facilitated the identification of pre-
data from patients undergoing ICB treatment in diverse cancers, existing chemoresistant genotypes within tumors prior to
researchers identified an elevation of tumor-specific neoadjuvant chemotherapy, thereby presenting potential oppor-
CXCL13+CD8+ T cells in responders (Liu et al., 2022b; Zhang tunities for predictive efficacy assessment before treatment (Kim
et al., 2021d). This past study also implied that CXCL13 can et al., 2018). In the context of radiochemotherapy (RCT), single-
serve as a potential unified biomarker for tumor-reactive CD8+ T cell sequencing has delineated the nuanced responses of specific
cells and ICB response. TME subpopulations, revealing features such as MHC-II upregu-
The TME encompasses a dynamic and intricate interplay lation in epithelial cells and an augmentation of naïve CD4+ cells,
among stromal cells, immune cells, and tumor cells, which CD16+ NK cells, and FCN1+ M-MDSC cells after RCT (Liu et al.,
plays a pivotal role in tumor progression and response or 2023a).
resistance to ICB therapy. By leveraging the high-resolution In recent years, the widespread adoption of combined
capabilities of single-cell analysis, researchers can discern immunotherapy with chemotherapy, radiotherapy, and other
distinct cellular subpopulations associated with immune modalities has demonstrated a notable enhancement in the
therapy resistance, thereby shedding light on strategies to overall response rates and survival rates, surpassing the out-
enhance ICB response rates or facilitate the development of comes achieved with individual approaches (Meric-Bernstam et
novel immunotherapeutic interventions. For instance, through al., 2021). Single-cell technology has advanced the exploration
single-cell profiling, two clusters, namely ecm-myCAF and of pertinent combination therapies, revealing diverse novel
TGFβ-myCAF, have been identified to interact with Tregs and combinations for various cancer types and facilitating the
exert influential roles in ICB resistance (Kieffer et al., 2020); in investigation of mechanisms that can contribute to enhanced
another investigation, the association between iCAFs and ICB therapeutic efficacy across malignancies such as NSCLC (Cascone
response was confirmed (Ma et al., 2023a). Notably, in gastric et al., 2023) and breast cancer (Bassez et al., 2021; Zhang et al.,
cancer, scRNA-seq analysis revealed the presence of fibroblast 2021d).

1260 SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 [Link]
Treatment of non-cancerous diseases virus of interest, thereby delineating disparities in immune
reactions to COVID-19 across diverse ethnicities and distinct
In addition to its applications in cancer treatment, single-cell cellular subpopulations (Aquino et al., 2023). Pioneering
genomics holds significant promise in the diagnosis and research utilizing scRNA-seq analysis has unraveled the intricate
treatment of a wide range of non-cancerous diseases, encom- interplay between COVID-19 and myeloid cells through C-type
passing viral inflammatory conditions, immunological diseases, lectin receptors and Tweety family member 2, which collectively
and neurological ailments. By leveraging the power of single-cell contribute to immune hyperactivation (Lu et al., 2021). More-
genomics, researchers can now unravel the intricate cellular over, the powerful combination of scBCR-seq and scRNA-seq
heterogeneity and functional diversity inherent in these diseases, techniques facilitates an effective identification of potent neu-
thereby shedding light on the underlying molecular mechanisms tralizing antibodies, thereby offering promising therapeutic
and identifying the potential therapeutic targets. This approach interventions and preventive strategies against prevalent in-
not only deepens our understanding of the complex cellular fectious diseases (Cao et al., 2020c).
landscapes associated with various non-cancerous ailments but In addition to the conventional disease therapies available,
also paves the way for the development of precision medicine single-cell technologies have exhibited a promising potential in
strategies. enhancing assisted reproductive technologies (ART) as a crucial
Autoimmune diseases, which encompass >80 distinct dis- tool in infertility treatment. The rapid advancement in single-cell
orders such as type 1 diabetes mellitus, inflammatory bowel multi-omics has facilitated the generation of high-resolution
disease, multiple sclerosis, rheumatoid arthritis, and systemic single-cell atlases of preimplantation human embryos (Li et al.,
lupus erythematosus, are characterized by dysregulated re- 2018), along with providing profound insights into the maternal-
sponses of B cells and T cells toward self-antigens, which triggers fetal interface interactions during the early gestation period
the generation of autoantibodies and self-reactive T cells (Vento-Tormo et al., 2018). These innovations have offered
(Pisetsky, 2023). The intricate nature of autoimmune diseases essential theoretical and technical foundations for the progres-
poses formidable challenges, with marked inter-patient hetero- sive evolution of ART. For instance, a clinical study on
geneity and a limited understanding of the underlying patho- preimplantation embryo DNA methylation profiling demon-
genic mechanisms (Harroud and Hafler, 2023). Single-cell strated that embryos with optimal methylation states are
genomics technologies offer valuable insights by enabling the associated with better clinical outcomes in ART (Gao et al.,
identification of TCR and BCR sequences from individual 2023).
lymphocytes, which empowers researchers to decipher the Regarding other prevalent complex diseases, the application of
intricacies of immune-related diseases and explore novel single-cell atlases collected for diverse disease tissues and organs
therapeutic avenues. Furthermore, the comprehensive profiling provides a heightened resolution for comparative analysis
of immune disorders using large-scale single-cell atlases, along between pathological and healthy tissues. These cell atlases
with the integration of healthy reference atlases, contributes proved instrumental in unveiling novel cellular subpopulations
significantly to the advancement of patient stratification and expressing disease-relevant genes, deciphering alterations in
precision medicine in autoimmune diseases (King et al., 2021; cellular composition, and elucidating intricate intercellular
Rood et al., 2022), such as in systemic lupus erythematosus crosstalk that underlies the pathogenic mechanisms across
(Arazi et al., 2019; Perez et al., 2022; Stewart et al., 2020), organs, such as in the liver (Aizarani et al., 2019), kidney
multiple sclerosis (Deng et al., 2021; Falcão et al., 2018), type 1 (Schreibing and Kramann, 2022), lung (Sikkema et al., 2023),
diabetes (Chiou et al., 2021; Hrovatin et al., 2023), Crohn’s brain (Garcia et al., 2022; Kanton et al., 2019; Winkler et al.,
disease (Mukherjee et al., 2023), and rheumatoid arthritis 2022), heart (Asp et al., 2019; Litviňuková et al., 2020), and
(Nathan et al., 2022; Yamada et al., 2023; Zhang et al., gastrointestinal tract (Fawkner-Corbett et al., 2021). Further-
2019b), as well as other immunological diseases such as more, the integration of multi-organ single-cell atlases facilitates
hypersensitivity syndrome (Kim et al., 2020). a comprehensive elucidation of the heterogeneity in the impact of
Many infectious diseases, particularly viral infections, elicit disease-associated genes on distinct tissues, as well as tissue-
infection-induced tissue inflammation, which causes immune specificity of the gene expression, cellular subtypes, and
dysregulation and profound clinical manifestations (Hill, 1998). compositional distributions (Elmentaite et al., 2022; Eraslan et
To comprehend the intricate immunopathogenesis underlying al., 2022; Mulder et al., 2021; The Tabula Sapiens Consortium et
these diseases, the utilization of large-scale single-cell atlases has al., 2022).
emerged as an invaluable tool, providing a comprehensive
depiction of tissue inflammation and enabling the identification
of distinct immune subsets (Rood et al., 2022). Notably, our
Drug discovery and precision medicine by single-cell
understanding of the immunological mechanisms governing
technology
certain infectious diseases remains limited. However, through the Target discovery: broadening the focus from cancer cells to holistic
application of single-cell genomics analysis, we can unravel the TME
complex interplay between pathogens and immune cells, thereby One of the key strategies in the pursuit of target discovery
elucidating disease mechanisms and unveiling novel therapeutic involves the direct targeting of driver genes in neoplastic cells.
targets. For instance, in the unprecedented global crisis of COVID- Preliminary analyses and screenings conducted through con-
19, the integration of an extensive single-cell landscape ventional genomic methods often lack the requisite granularity,
encompassing various disease severities and stages uncovered necessitating substantial experimental efforts and temporal
an array of immune characteristics associated with COVID-19 investments to obtain an expansive list of potential targets.
infection (Ren et al., 2021). These atlases also furnish intricate Single-cell technologies offer an efficacious approach in this
details about the cellular and molecular responses elicited by the direction. For instance, leveraging scRNA-seq and scDNA-seq for

[Link] SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 1261
intratumor heterogeneity dissection and the analysis of epithe- throughput methodologies empower the detection of the func-
lial-mesenchymal transition (EMT)-related genes is a strategic tional impact of multiple targeted genes across cellular subtypes,
screening approach for tumor cell genes, which facilitates the enabling streamlined and accurate screening of candidate
discovery of new targets and innovative treatment strategies targets. To further alleviate the experimental burden, several
(Gavish et al., 2023; Pellegrino et al., 2018). Within the deep learning frameworks leveraging single-cell perturbation
spectrum of other pathological conditions, the convergence of data have been developed. Models like GEARS, which incorporate
genome-wide association studies (GWAS) with single-cell ex- graph neural networks, predict the downstream effects of
pression quantitative trait loci (sc-eQTL) (Yazar et al., 2022), multiple gene perturbations, even those not experimentally
STING-seq (Morris et al., 2023), sc-linker framework (Jagadeesh perturbed, thereby guiding experimental design and mitigating
et al., 2022), and other singular cell sequencing modalities costs (Roohani et al., 2024). Another approach, known as
facilitates the comprehensive integration of RNA sequences and CINEMA-OT, employs causal independent effect module attribu-
human genetics. By delving into the molecular intricacies of tion based on causal inference to undertake single-cell perturba-
GWAS mutations, this approach probes into the causal relation- tion analysis and prognosticate therapeutic efficacy (Dong et al.,
ships among cell types, risk loci, and the associated diseases, 2023).
thereby elucidating potential therapeutic target genes or cells. Following the identification of potential drug candidates, the
Therapeutic approaches exclusively targeting cancer cells may subsequent steps involve high-throughput screening (HTS) of
encounter challenges in accessing certain targets, as evidenced drug reagents. Conventional HTS methods, though effective, are
by the limited efficacy of CAR-T in solid tumors. The burgeoning often associated with high costs and significant time investments,
field of single-cell genomics, when coupled with advancements in as each screening step can only be conducted using a single
integrative analyses, proffers a novel paradigm for target dosage and under certain specific biological conditions (Macar-
discovery, which pivots on the identification of therapeutically ron et al., 2011). The advent of single-cell technology, when
relevant subtypes of TME cells through single-cell omics so as to integrated with HTS, allows for simultaneous testing under
uncover pertinent genes and modulation of their cellular multiple dosages and diverse conditions (Van de Sande et al.,
interactions, consequently skewing the balance in favor of a 2023). Moreover, a range of new single-cell methods have
tumoricidal TME. This approach encompasses functional aug- emerged to enhance the cost-effectiveness and efficiency of HTS
mentation or recruitment of anti-tumorigenic cellular subtypes in combination with scRNA-seq, either by reducing the cost or
such as CXCL13+ precursors of exhausted CD8+ T cells (Liu et al., increasing the throughput. For example, the Sci-Plex method,
2022b), CXCL13+CD4+ T cells (Zhang et al., 2021d), and the which employs Nuclear hashing and single-cell combinatorial
diminution of pro-tumorigenic cellular subtypes like activated indexing sequencing (sci-RNA-seq), can capture the global
Treg (Guo et al., 2018), SPP1+ macrophages (Zhang et al., transcriptional response of cancer cells to thousands of indepen-
2020a), TaNK (Tang et al., 2023a), LRRC15+ CAFs (Krishna- dent perturbations at a single-cell resolution within a single
murty et al., 2022), or the obstruction of their interactive experiment (Srivatsan et al., 2020).
pathways, potentially offering a more holistic and nuanced
approach to neoplastic therapy. Patient stratification and personalized precision therapy
Furthermore, various cell therapies can be harnessed across a The intricate heterogeneity observed among distinct tumors
spectrum of diseases including cancer (Weber et al., 2020), and necessitates the precise stratification of different cancer patients
the construction of a tumor antigen map through single-cell during therapy selection, as it is a critical step in enhancing the
genomics furnishes potential targets for diverse cell therapies in efficacy and safety of cancer treatment. Current studies on tumor
cancer (Lareau et al., 2021b). Methods affiliated with single-cell stratification, which rely on conventional genomics approaches,
genomics can thus foster the development of diverse cell or are primarily based on genetic or epigenetic profiles as well as
immune therapies, including deciphering of biological character- alterations in oncogenic pathways (Guinney et al., 2015;
istics and adverse reaction mechanisms of CAR-T and unearthing Hoadley et al., 2018). The use of enhanced resolution in single-
potential novel strategies via single-cell technologies and cell studies, especially in analyses of large-scale data, enables
methodologies such as scRNA-seq, scTCR-seq, scATAC-seq, better-refined molecular stratification of cancer patients, which
CyTOF, and CITE-seq (Huang et al., 2023b). incorporates not only cancer cells but also immune cells in the
TME, thereby guiding better targeting strategies for accurate,
Single-cell high-throughput target and drug screening precise, and personalized immunotherapy. For instance, a
The potential targets identified through preliminary analysis classification of pan-cancer samples based on the composition
necessitate further screening to ascertain their functional of tumor-infiltrating T cells from a large-scale pan-cancer
relevance to certain disease biology via genetically encoded landscape of T cells can have distinct clinical implications and
perturbations or preclinical models. Perturbations induced by preferences for certain cancer types. These findings suggest that
specific candidate genes often elicit distinct transcriptional patient groups of cancer types featured with a high proportion of
changes across heterogeneous cellular subpopulations or tissue exhausted T cells and low resident memory T cells have worse
organs. Conventional methodologies, such as the use of a overall survival and that cancer types enriched in the group of
combination of pooled CRISPR screening approach with bulk high numbers of exhausted T cells and low Treg, such as
RNA-seq, may obscure the nuanced functionality of these genes melanoma, are indicative of better response to ICB therapy
in minority cell populations (Bock et al., 2022). To surmount this (Zheng et al., 2021). T cell-based patient stratifications are
limitation, emerging scCRISPR-seq such as perturb-seq (Dixit et associated with the efficacies of T cell-targeted therapy; as such,
al., 2016), which integrates with scRNA-seq, as well as ECCITE- the stratifications considering integration of other TME cells
seq, which has expanded CRISPR to multimodal single-cell might provide a broader perspective for deciphering the complex-
screens (Papalexi et al., 2021), have been devised. These high- ities in therapy selection for immunotherapy targeting a wider

1262 SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 [Link]
range of cells (Joanito et al., 2022; Leader et al., 2021). In approach with broad implications in cancer therapeutics as well
addition, the intricate nature of dynamic cellular interactions as in the treatment of other complex diseases. By broadening the
within the TME necessitates surpassing the limitations of investigative perspective, such as dissecting TME and patient
categorizing single-cell types or individual cancer types, thereby stratification, single-cell genomics can facilitate drug target
pressing for further leaps in accurate patient stratifications based discovery and screening, thereby propelling the efficacy of
on large-scale single-cell integration of all TME components precision medicine. The ongoing exploration of the vast potential
across numerous cancer types, which can ultimately guide the of single-cell genomics within the realm of healthcare holds
personalized selection of immunotherapy regimens. significant promise toward heralding imminent transformative
The challenge posed by pan-essential cancer targets may changes in clinical practices.
engender inadvertent toxicities in patients (Chang et al., 2021).
Patient stratification through single-cell genomics, hinging on
Chapter 8. Perspectives, challenges, and
the distinct enrichment profiles of diverse cell subtypes within
opportunities of single-cell genomics
patients, can allow the implementation of patient-specific
targeted pharmacotherapies or treatment modalities, whose With synergistic advances made in experimental and computa-
variance can be comparatively assessed by leveraging identified tional techniques in the fields of single-cell and spatial genomics,
biomarkers. Patient stratification paradigms have now extended it has been revealed, in the context of both physiology and
beyond the realm of cancer and have found applications in pathology, how heterogeneous the transcriptome of individual
diverse diseases. In Crohn’s disease, for example, scRNA-seq has cells can be in a seemingly homogeneous cell population (Kelsey
unveiled the cellular module termed GIMATS, which is et al., 2017; Lein et al., 2017; Stubbington et al., 2017). Despite
intricately linked to anti-TNF therapy resistance (Martin et al., these advances, several challenges remain to be resolved in the
2019). Similarly, in human chronic inflammatory skin disease, a successful application of single-cell genomics in the future. First,
sophisticated tool that leverages single-cell characteristics has single-cell multi-omics has become an indispensable approach to
been devised for the purposeful classification of rash samples (Liu elucidate the complexity of cellular processes. Although numer-
et al., 2022d). ous techniques have been developed so far, there is much scope
for improvement in the protocols to address issues such as limited
Deciphering adverse effects of immunotherapy coverage so as to capture more informative and complete
In recent years, the landscape of cancer treatment has been genomic data (Baysoy et al., 2023). Second, spatial genomics
significantly altered by the advent of immunotherapeutic has revolutionized our current understanding of cellular
strategies, with ICB standing out as a pivotal player therein. processes by providing the exact location of cells. Although
However, the clinical efficacy of these approaches is intricately several spatially resolved single-cell multi-omics techniques have
entwined with the emergence of immune-related adverse events been developed so far, most of these methods are low-throughput
(irAEs), which affect a diverse array of organs, including the and have low sensitivity, and hence cannot capture the complete
gastrointestinal tract, skin, liver, kidneys, and lungs, thereby spatial architecture of tissue with most of the molecular
introducing a formidable challenge in patient care and manage- information (Baysoy et al., 2023). Third, the cost of single-cell
ment (Martins et al., 2019). Despite the generally lower incidence spatial genomics is very high, and there is a need to develop cost-
of irAEs in comparison to that in conventional treatments, their effective and scalable methods for their routine application.
nuanced and multifaceted characteristics impose substantial Another important challenge in this area arises from data
hurdles in the realm of clinical management, which, to some integration. Matched data from single-cell multi-omics are
extent, constrain the progressive advancement of immunothera- relatively easy to handle since distinct modalities can be tethered
pies (Gangadhar and Vonderheide, 2014). The application of to a single cell. However, future improvements are required for
single-cell sequencing and analysis has emerged as a pivotal the integration of unmatched data, considering the excessive
approach for determining the intricate mechanisms of irAEs and single-cell data available with singular modality that has been
for propelling the mitigation of adverse side effects. Single-cell generated over the past decade (Argelaguet et al., 2021). To
genomics research has elucidated the pivotal role of α-myosin as move forward to successful clinical translation, there is a need to
a significant autoantigen in immune checkpoint inhibitor- bridge the gap between research and clinical application for
associated myocarditis (ICI-MC), which acts as a potent stimulus improving patient outcomes. Finally, since single-cell and spatial
for clonal T-cell expansion and incites autoimmune inflamma- genomics methods amass huge amounts of genomic information,
tory responses (Axelrod et al., 2022). Notably, peripheral α- ethical issues related to data privacy and application in clinical
myosin-specific T-cells have been identified as a promising settings would also need to be considered.
biomarker for predicting myocarditis risk. Another study high-
lighted the enrichment of Temra CD8+ cells in patients with ICI-
Increase data coverage
MC and an upregulation of chemokines CCL5/CCL4/CCL4L2
(Zhu et al., 2022). In checkpoint inhibitor-induced colitis (+CPI Recent genomic technologies for studying single-cell gene
colitis), comprehensive single-cell analysis of colitis data eluci- expression (scRNA-seq), chromatin accessibility (scATAC-seq),
dated a pronounced proliferation and accumulation of cytotoxic and chromatin interactions (scHi-C, Dip-C) have tremendously
effector CD8+ T cells, indicating their potential implication in the advanced our understanding of cellular function and diversity.
pathogenesis of +CPI colitis while also identifying several However, there is a major obstacle in this direction—data
cytokines, chemokines, and surface receptors as prospective sparsity. Although the current single-cell data coverage is
therapeutic targets for any associated side effects (Luoma et al., generally adequate for identifying cell types and clustering, it
2020). lacks the depth needed for a thorough characterization of
Single-cell genomics has thus emerged as a transformative individual cells. The additional data required include their

[Link] SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 1263
intricate properties, specific functions, response to external protein. The traditional mono-omics methods can not reveal the
stimuli and drugs, and their differentiation or reprogramming direct causality between different types of molecular information.
paths. To fully harness the potential of these technologies and to Multi-omics technologies, however, represent a pivotal advance-
aspire toward creating cellular reference maps of organisms and ment in molecular biology that is designed to capture multiple
detailing the position, function, and characteristics of every types of molecular information within the same biological
individual cell, the coverage of single-cell data must be increased samples by a single experiment. To understand the molecular
significantly. hierarchy of genome structures and functions, methods like
In scRNA-seq, the utilization of random priming and in situ Chromatin Interaction Analysis with Paired-End Tag (ChIA-PET)
polyadenylation have enabled full-length coverage and detection (Fullwood et al., 2009) have been developed to detect TF binding
of total RNA (McKellar et al., 2023; Salmen et al., 2022; Xu et sites and associated chromatin interactions, while Chromatin
al., 2023f). In addition, the application of diverse single-molecule Interaction Analysis with ATAC-seq (ChIATAC) (Chai et al.,
labeling strategies is poised to elevate the detectable spectrum of 2023) was created to simultaneously measure chromatin
RNA (Mahat et al., 2024; Wu et al., 2024b; Yao et al., 2023; Ye accessibility and interactions in cell populations. However,
et al., 2024b). These innovative methods can extend their utility despite the great strides accomplished and the known potential,
to STs or in situ sequencing, thereby enhancing the sensitivity of two primary challenges persist. First, Hi-C-related data is
single-cell and spatial genomic methods. inherent to limited resolution, which hampers the detailed
In sn-ATAC-seq, different strategies may have the potential to characterization of interactions between cis-regulatory elements
increase the sensitivity. For instance, Tn5 transposase has been and their target gene promoters. Second, multi-omics single-cell
applied to insert T7 promoters (T7P) into genomic DNA (gDNA) studies involving 3D genome analysis have limited throughput.
to facilitate in vitro linear amplification and thereby reduce PCR- Particularly, the latter represents a serious bottleneck, constrain-
introduced bias during exponential amplification (Bartlett et al., ing our ability to perform comprehensive and detailed studies at
2021; Chen et al., 2017a; Lake et al., 2018; Sos et al., 2016). scale and, thus, slowing the pace of gathering insights into the
More recently, Mulqueen et al. implemented a method using a 3D genomic organization across a diverse range of cell types and
single-adaptor transposition strategy (Mulqueen et al., 2021) conditions.
called s3-ATAC that utilizes a forward-adaptor containing a The next frontier in multi-omics research lies in overcoming
uracil base. By employing a uracil-intolerant polymerase for gap- current limitations to achieve high-resolution mapping of
filling, this approach prevents the incorporation of a forward chromatin interactions, particularly to gene-centric contacts
adapter sequence on both ends. Subsequently, the reverse involving specific promoter-enhancer interactions that directly
adapter is introduced with a polymerase-based adaptor switch. modulate the gene expression for determining cell identity. By
In this step, a reverse primer with locked nucleic acid is combining this approach with RNA expression information
incorporated into the template by using a uracil-processing within the same cell, researchers can investigate the relationship
polymerase. In addition, the adoption of long-read strategies in between the genome structure with transcription. Additionally,
single-cell analyses, such as scNanoHi-C (Li et al., 2023c), has incorporating protein information may provide mechanistic
shown promise, demonstrating approximately nearly complete insights that span from DNA to RNA to protein synthesis—a
coverage of genomic regions (99.3% of 100-kb pair genomic process that is at the heart of the central dogma of molecular
bins). biology.
In spatial and in situ sequencing, 10x Genomics Xenium in situ Equally critical is tackling the chemical incompatibilities
gene expression profiling has expanded the target panel from 300 inherent in multi-modal methodologies, which often lead to
to 5,000 genes for single-cell spatial imaging (https:// reduced detection capabilities when compared with that with
[Link]/platforms/xenium). Deep-STARmap single-modality analyses. In these directions, it is foreseeable that
and Deep-RIBOmap enabled single-cell in situ transcriptional additional modalities of molecular information can be included in
and translational gene expression of 200-μm thick tissue blocks the future multi-omics single-cell methods for simultaneously
(Sui et al., 2024). Besides single-molecular DNA sequencing mapping RNA expression and interactions among DNA-DNA,
(Pacbio, Nanopore) and smFISH-based strategy (Vizgen, Nano- DNA-proteins (e.g., histone and TF), DNA-RNA, and RNA-RNA.
String, 10x Genomics Xenium), single-molecule protein sequen- Such multi-omics molecular information derived in the same cell
cing has emerged as a promising way for integrated omics (Reed will greatly enhance our ability to establish the causal relation-
et al., 2022). Subtle cellular structure analyzing methods for ship between 3D epigenomic states to their functional outcomes
nuclei, mitochondria (Guo et al., 2023), and ribosome (VanIns- in transcription regulation and cellular functions. To support
berghe et al., 2021; Zeng et al., 2023a) provides more elaborate widespread, large-scale applications of multi-omics research,
information about cytogenetics and cell kinetics. systematic comparison and standardization of multiple metho-
Cumulatively, these innovations have significantly improved dological variations will be essential. Major projects such as
the complexity of libraries for individual cells, thereby enhancing ENCODE, BICCN, and 4DN have used multi-omics technologies to
detection sensitivity and disclosing subtle cellular variations. build comprehensive atlases (BRAIN Initiative Cell Census
Advancements along these lines are, therefore, poised to further Network (BICCN), 2021; Dekker et al., 2017; Peng et al., 2021).
deepen our insights into the myriad of biological intricacies Multi-modal datasets can act as a “molecular” bridge (Hao et
inherent in a single cell. al., 2024b) for connecting single-cell datasets across modalities.
Although the existing mono-modality single-cell research has
provided a deep dive into specific biological aspects, its insights
Increase multi-omics of single-cell mapping
remain isolated, which is limited to the scope of each modality.
Cellular mechanisms of biological functions involve the central Encompassing information from disparate modalities, multi-
dogma of genetic information converting from DNA to RNA to modal datasets can enable effective translation and integration

1264 SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 [Link]
within these modalities, free from the reliance on constraining tion, ease-of-use, and cost-efficient open-source spatial omics
biological presumptions. In summary, advancements in these frameworks are expected to lead to the widespread application of
directions will not only refine our current understanding of the these technologies in medical organizations (Schott et al., 2024).
cellular machinery but also enhance our ability to decipher the
molecular underpinnings of a healthy state and a diseased state.
AI for single-cell science
On one hand, single-cell technologies produce highly systematic
Integrate single-cell multi-omics with spatial omics
data, which makes them well-suited for AI training. On the other
For multi-cellular organisms, cellular function is highly depen- hand, AI has been providing revolutionary applications in single-
dent on the context of spatial localization within a given tissue or cell genomics, thereby elevating the field to new heights. It has
organ. Although recent advancements in imaging and sequence- revolutionized data preprocessing, batch-effect correction, data
based spatial technologies have revolutionized our understand- normalization, dimension reduction, feature selection, cell
ing of how gene expression patterns are influenced by the spatial clustering, cell type annotation, graphic presentation, and other
arrangement of cells within tissues, spatial omics technologies aspects. We have provided here an overview of the broad scope of
are currently constrained by limited capturing area and low AI applications in single-cell science, including (i) advanced deep
chemistry efficiency. As a result, spatial omics data have a learning models (Chen et al., 2022c; de la Fuente et al., 2023;
reduced capacity to capture genomic and transcriptomic features Menden et al., 2020) that enable deconvolution from bulk to
when compared with single-cell analyses. To capitalize on the single-cell data, thereby capturing individual cell behaviors with
existing single-cell consortia data (Regev et al., 2017), efforts precision; (ii) AI methods (Amodio et al., 2019; Arisdakessian et
have been made to integrate the scRNA-seq data with the ST al., 2019; Deng et al., 2019; Eraslan et al., 2019; Lopez et al.,
data. Despite these efforts, the existing computational tools for 2018) can represent cells in a low-dimensional linear space or
integration often lack precision; they typically estimate the cell latent form to fill in any missing or unseen gene expression
types within each spatial spot rather than obtaining a single-cell values, thereby enhancing the overall data quality; (iii) AI-driven
resolution (Wan et al., 2023). Advancing computational tools is algorithms like SAUCIE (Amodio et al., 2019), scGen (Lotfollahi
critical in enhancing single-cell analyses, as it enables research- et al., 2019), and scVI (Lopez et al., 2018) correct batch effects to
ers to fully leverage spatial context toward understanding align datasets from different experimental conditions; (iv) by
cellular functions and interactions. This development is key to utilizing techniques such as GPTCelltype (Hou and Ji, 2024), AI
unlocking the full potential of spatially informed single-cell automatically classifies or labels different cell types based on the
biology. gene expression profiles; (v) AI models like GLUE (Cao and Gao,
Recently, Russell et al. (2024) introduced a groundbreaking 2022) incorporate multiple layers of biological data to provide
method that involved tagging tissues with spatial barcodes, deeper insights into cellular processes; (vi) AI models like CAME
followed by the recovery of the desired nuclei for robust single- (Liu et al., 2023c) and SATURN (Rosen et al., 2024) integrate
cell RNA-seq analysis. These spatial barcodes are ingeniously datasets across multiple species to understand cell type evolution
designed to integrate seamlessly with downstream single-cell and conservation; (vii) AI methods like DeepST (Xu et al.,
processing workflows. The spatially barcoded nuclei can then be 2022a), STAGATE (Dong and Zhang, 2022), and DeLTA2.0
input into commercially available single-cell microfluidic plat- (O’Connor et al., 2022) capture both the gene expression profiles
forms, thereby bridging the single-cell measurements with spatial and spatial information, thereby enriching our current under-
genomics insights. Although the Slide-tag technique currently standing of cellular organization within tissues; (viii) AI
faces challenges due to sample loss, which limits analysis to only techniques such as DeepCCI (Yang et al., 2023a) enable the
a subset of the tissue, it represents a significant step forward in comprehensive exploration of interaction between different cell
linking single-cell analysis with spatial context, thereby enhan- types to reveal complex cellular dynamics; (ix) AI techniques like
cing our understanding of cellular functions within their native DeepVelo (Chen et al., 2022e) enable reconstruction of temporal
environments. trajectories of cellular states, thereby unraveling dynamic
Alternatively, an appealing strategy to link spatial omics data processes such as differentiation and development; (x) AI models
with single-cell multi-omics data, or conversely, to assign single- such as Nvwa (Li et al., 2022c) decodes the secrets of regulatory
cell multi-omics information with spatial localization in the tissue grammars, which plays a pivotal role in elucidating core
context, is through computational integration. For matched ST regulatory mechanisms governing gene expression and provid-
data and single-cell multi-omics datasets with RNA, 3D genome, ing valuable insights into cell fate determination; (xi) pre-trained
and epigenome modalities from the same biological sample, the models like Geneformer has an impressive success in the context
shared spatial RNA and the single-cell RNA data can function as of gene regulatory network inference in the field of single-cell
an information bridge to connect tissue anatomic location with science. Overall, AI is shaping single-cell data analysis (Ma and
the 3D epigenomic information. Therefore, future development in Xu, 2022).
this direction by leveraging the combined capabilities of single- AI models have become a powerful tool in the field of spatial
cell precision and spatial resolution is positioned at the cutting data, with their applications ranging from cell type identification
edge of scientific discovery. Technically, future directions for to functional perdition. These applications leverage AI’s ability to
spatial omics may involve click chemistry, chemical labeling, and analyze complex spatial patterns and relationships within
a computational-based “optics-free” system for single-molecular biological systems. The key areas where AI can significantly
imaging reconstruction (Weinstein et al., 2019). The application contribute include (i) models like DeepST and STAGATE leverage
foreground of spatial omics may focus on the 3D construction of a graph-based methods to reconstruct spatial relationships, which
large number of clinical samples (especially formalin-fixed provides a more comprehensive understanding of spatial
paraffin-embedded tissues) for biomarker detection. High-resolu- organization; (ii) AI algorithms like DSTG (Song and Su, 2021)

[Link] SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 1265
and GTAD (Zhang et al., 2023c) utilize graph-based convolu- be transformative in developing targeted therapies and bridging
tional network to perform spot deconvolution, parsing spot-based the gap between research and clinical applications for improved
data into single-cell level expression profiles; (iii) models such as patient-related clinical outcomes.
SpaGCN (Hu et al., 2021) and GLISS (Zhu and Sabatti, 2020) Previous studies have suggested that the majority of variants
integrate gene expression, spatial location, and histology to identified through GWAS are situated in the extensive non-
identify spatial domains and spatially variable genes through the coding regions of the human genome, which comprises
application of graph convolutional networks; (iv) models like approximately 91% of the total genome (Claringbould and
GCNG and HoloNet are commonly employed to infer cell-cell Zaugg, 2021). These non-coding regions do not directly affect
interactions at a spatial resolution. Notably, graph-based models protein coding, implying an enhancer-mediated regulatory role.
have emerged as a powerful tool for spatial data analysis. The However, GWAS studies often fail to link variants to specific
ability to represent complex spatial relationships within tissues as target genes and do not specify the functions of these variants in
graphical topological structures has unlocked new possibilities specific cell types or tissues. As such, linking GWAS variants to
for understanding tissue organization and function to greater their target genes is crucial for developing targeted therapies.
depths. Single-cell multi-omics 3D genome mapping technologies can
Future AI models within the field of single-cell science are enable precise mapping of enhancers at single-cell resolution (Liu
expected to leverage multi-omics data even further so as to et al., 2023f). This precise mapping will, in turn, facilitate the
enhance our understanding of gene regulatory rules and syntax. connection of GWAS loci with their target genes, thereby
The evolution of AI architectures from CNN-based to attention- enabling the development of drugs targeting these genes and
based models is a testament to the power of the latter in dealing allowing the enhancers to serve as direct targets for putative
with complex, long-range relationships. Current specialized therapy. For example, in diseases where enhancers drive the
transformer models like scGPT, scMoFormer (Tang et al., expression of undesirable gene products, editing the enhancer
2023b), and scCLIP (Xiong et al., 2023) have been constructed sequence via CRISPR editing techniques presents a promising
for single-cell omics prediction. More importantly, systems treatment option. This therapy has been pioneered in transfu-
biology will witness considerable benefits from AI models inspired sion-dependent β-thalassemia and sickle cell disease (Frangoul et
by big data, which will shift it from a mere integrative subject to al., 2021). Unlike targeting specific proteins or enzymes,
one with heightened predictive capabilities. targeting disease-related enhancers offers greater cell type
specificity and tissue specificity, with the promise of minimizing
the disruption of gene functions of other cell types.
Clinical translations
In clinical practice, the application of single-cell genomics can
In addition to their profound impact on our current under- benefit the oncological, hematological, and immunological
standing of disease pathogenesis, drug discovery, and precision aspects of diagnostic biomarkers and therapeutic target detection
medicine, single-cell and spatial genomics have the potential to (Lim et al., 2023), especially for epigenetic targeted therapy

Figure 10. Future directions in single-cell and spatial genomics. A, Increasing data coverage in single cells (NGS, next-generation sequencing; SMRT, single-molecule real-time
sequencing). B, Enabling of multi-modal detection in single cells. C, Integration of single-cell genomic data with spatial genomic and in situ sequencing data. D, Leveraging AI for
single-cell and spatial genomics.

1266 SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 [Link]
(Yamagishi et al., 2024). Multimodal single-cell genomics Supporting information
The supporting information is available online at [Link]
approaches have enabled wholescale sequencing of small, frozen The supporting materials are published as submitted, without typesetting or editing. The
clinical specimens (Wang et al., 2023c). Spatial genomics of responsibility for scientific accuracy and content remains entirely with the authors.
clinical FFPE samples can further provide a retrospective
pathology landscape for tumor treatment (Jackson et al., References
2020). Technical advances in single-cell and spatial genomics Adil, A., Kumar, V., Jan, A.T., and Asger, M. (2021). Single-cell transcriptomics:
have provided solutions for different clinical samples. The cell- current methods and challenges in data acquisition and analysis. Front Neurosci
15, 591122.
type specific regulatory events in disease can inspire cutting-edge Ai, S., Xiong, H., Li, C.C., Luo, Y., Shi, Q., Liu, Y., Yu, X., Li, C., and He, A. (2019).
therapies, including genome editing, oncolytic virus (Ling et al., Profiling chromatin states using single-cell itChIP-seq. Nat Cell Biol 21, 1164–
2023), cell and immunotherapy, and epigenetic targeted small 1172.
molecules. Aizarani, N., Saviano, A., Sagar, A., Mailly, L., Durand, S., Herman, J.S., Pessaux, P.,
Baumert, T.F., and Grün, D. (2019). A human liver cell atlas reveals heterogeneity
In the forthcoming years, the advent of single-cell and spatial
and epithelial progenitors. Nature 572, 199–204.
genomics is poised to facilitate the establishment of the Almanzar, N., Antony, J., Baghel, A.S., Bakerman, I., Bansal, I., Barres, B.A., Beachy,
regulatory atlas, detailing cell type-specificity and the develop- P.A., Berdnik, D., Bilen, B., Brownfield, D., et al. (2020). A single-cell
mental stages’ specificity. Fine-tuning the activity of regulators transcriptomic atlas characterizes ageing tissues in the mouse. Nature 583, 590–
or regulatory elements can empower clinicians to achieve precise 595.
Alon, S., Goodwin, D.R., Sinha, A., Wassie, A.T., Chen, F., Daugharthy, E.R., Bando,
control over gene expression, thereby expanding the possibilities Y., Kajita, A., Xue, A.G., Marrett, K., et al. (2021). Expansion sequencing: spatially
for targeted treatments. This approach may emphasize a broader precise in situ transcriptomics in intact biological systems. Science 371, eaax2656.
regulatory genomic landscape, thereby laying a novel ground- Amaral, P., Carbonell-Sala, S., De La Vega, F.M., Faial, T., Frankish, A., Gingeras, T.,
work for future drug development. Guigo, R., Harrow, J.L., Hatzigeorgiou, A.G., Johnson, R., et al. (2023). The status
of the human gene catalogue. Nature 622, 41–47.
Amodio, M., van Dijk, D., Srinivasan, K., Chen, W.S., Mohsen, H., Moon, K.R.,
Conclusion Campbell, A., Zhao, Y., Wang, X., Venkataswamy, M., et al. (2019). Exploring
single-cell data with deep multitasking neural networks. Nat Methods 16, 1139–
With improvements in throughput, sensitivity, and modalities, 1145.
single-cell and spatial technologies are expected to drive a Anchang, B., Hart, T.D.P., Bendall, S.C., Qiu, P., Bjornson, Z., Linderman, M., Nolan,
G.P., and Plevritis, S.K. (2016). Visualization and cellular hierarchy inference of
paradigm shift in all facets of biological investigation. In addition,
single-cell data using SPADE. Nat Protoc 11, 1264–1279.
the synergistic integration of AI stands as a cornerstone, poised to Angerer, P., Simon, L., Tritschler, S., Wolf, F.A., Fischer, D., and Theis, F.J. (2017).
bolster these efforts to enable more advanced analysis and Single cells make big data: new challenges and opportunities in transcriptomics.
interpretation of genomic data (Figure 10). Together, these Curr Opin Syst Biol 4, 85–91.
innovations promise to usher in a new era of discovery, with the Angermueller, C., Clark, S.J., Lee, H.J., Macaulay, I.C., Teng, M.J., Hu, T.X., Krueger,
F., Smallwood, S.A., Ponting, C.P., Voet, T., et al. (2016). Parallel single-cell
potential to reveal new biological insights and translate them sequencing links transcriptional and epigenetic heterogeneity. Nat Methods 13,
into clinical applications for improved human health and disease 229–232.
management. Aquino, Y., Bisiaux, A., Li, Z., O’Neill, M., Mendoza-Revilla, J., Merkling, S.H., Kerner,
G., Hasan, M., Libri, V., Bondet, V., et al. (2023). Dissecting human population
variation in single-cell responses to SARS-CoV-2. Nature 621, 120–128.
Summary Aran, D., Hu, Z., and Butte, A.J. (2017). xCell: Digitally portraying the tissue cellular
heterogeneity landscape. Genome Biol 18, 220.
Recent advancements in single-cell and spatial genomics have Aran, D., Looney, A.P., Liu, L., Wu, E., Fong, V., Hsu, A., Chak, S., Naikawadi, R.P.,
revolutionized the research on cellular states and heterogeneity. Wolters, P.J., Abate, A.R., et al. (2019). Reference-based analysis of lung single-cell
Innovations in single-cell multi-omics and spatial genomics have sequencing reveals a transitional profibrotic macrophage. Nat Immunol 20, 163–
172.
opened up ways to integrate molecular data with the mapping of Arazi, A., Rao, D.A., Berthier, C.C., Davidson, A., Liu, Y., Hoover, P.J., Chicoine, A.,
cellular features within tissue contexts. These technologies have Eisenhaure, T.M., Jonsson, A.H., Li, S., et al. (2019). The immune cell landscape in
enabled the construction of detailed cellular atlases, offering new kidneys of patients with lupus nephritis. Nat Immunol 20, 902–914.
insights into tissue organization and disease mechanisms. In Arendt, D., Musser, J.M., Baker, C.V.H., Bergman, A., Cepko, C., Erwin, D.H., Pavlicev,
addition, improved computational tools have enhanced data M., Schlosser, G., Widder, S., Laubichler, M.D., et al. (2016). The origin and
evolution of cell types. Nat Rev Genet 17, 744–757.
integration, while systematic data from single-cell and spatial Argelaguet, R., Cuomo, A.S.E., Stegle, O., and Marioni, J.C. (2021). Computational
genomics have paved the way for digital cell atlases across principles and challenges in single-cell data integration. Nat Biotechnol 39, 1202–
various systems. Importantly, AI-driven predictive foundation 1215.
models are emerging to enable in silico perturbation screening Arisdakessian, C., Poirion, O., Yunits, B., Zhu, X., and Garmire, L.X. (2019).
DeepImpute: An accurate, fast, and scalable deep neural network method to impute
and synthetic regulatory network design to advance our current
single-cell RNA-seq data. Genome Biol 20, 211.
understanding of complex diseases and guide targeted therapeu- Arrastia, M.V., Jachowicz, J.W., Ollikainen, N., Curtis, M.S., Lai, C., Quinodoz, S.A.,
tic strategies. Selck, D.A., Ismagilov, R.F., and Guttman, M. (2022). Single-cell measurement of
higher-order 3D genome organization with scSPRITE. Nat Biotechnol 40, 64–73.
Compliance and ethics Arutyunyan, A., Roberts, K., Troulé, K., Wong, F.C.K., Sheridan, M.A., Kats, I.,
The authors declare that they have no conflict of interest. Garcia-Alonso, L., Velten, B., Hoo, R., Ruiz-Morales, E.R., et al. (2023). Spatial
multiomics map of trophoblast development in early pregnancy. Nature 616, 143–
Acknowledgement 151.
This work was supported by the Key R&D Program of Zhejiang (2024SSYS0022), the National Ashuach, T., Gabitto, M.I., Koodli, R.V., Saldi, G.A., Jordan, M.I., and Yosef, N.
Natural Science Foundation of China (31930028, 91842301, 32370608, 32250710678, (2023). MultiVI: Deep generative model for the integration of multimodal data. Nat
32200073, 32250710678), Leading Innovative and Entrepreneur Team Introduction Program Methods 20, 1222–1231.
of Zhejiang (2021R01012), “Pioneer” R&D programs of Zhejiang Province (2024C03005), Asp, M., Giacomello, S., Larsson, L., Wu, C., Fürth, D., Qian, X., Wärdell, E., Custodio,
STI2030-Major Projects (2021ZD0202502), the National Key Research and Development J., Reimegård, J., Salmén, F., et al. (2019). A spatiotemporal organ-wide gene
Program of China (2022YFC3400400, 2018YFA0800503, 2018YFA0107804) and Beijing expression and cell atlas of the developing human heart. Cell 179, 1647–1660.
Municipal Science and Technology Commission (Z221100007022002). e19.

[Link] SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 1267
Athaya, T., Ripan, R.C., Li, X., and Hu, H. (2023). Multimodal deep learning using EEL FISH. Nat Biotechnol 41, 222–231.
approaches for single-cell multi-omics data integration. Brief Bioinf 24, bbad313. BRAIN Initiative Cell Census Network (BICCN) (2021). A multimodal cell census and
Avsec, Ž., Agarwal, V., Visentin, D., Ledsam, J.R., Grabska-Barwinska, A., Taylor, K. atlas of the mammalian primary motor cortex. Nature 598, 86–102.
R., Assael, Y., Jumper, J., Kohli, P., and Kelley, D.R. (2021). Effective gene Braun, E., Danan-Gotthold, M., Borm, L.E., Lee, K.W., Vinsland, E., Lönnerberg, P.,
expression prediction from sequence by integrating long-range interactions. Nat Hu, L., Li, X., He, X., Andrusivová, Ž., et al. (2023). Comprehensive cell atlas of the
Methods 18, 1196–1203. first-trimester developing human brain. Science 382, eadf1226.
Axelrod, M.L., Meijers, W.C., Screever, E.M., Qin, J., Carroll, M.G., Sun, X., Tannous, Bravo González-Blas, C., Minnoye, L., Papasokrati, D., Aibar, S., Hulselmans, G.,
E., Zhang, Y., Sugiura, A., Taylor, B.C., et al. (2022). T cells specific for α-myosin Christiaens, V., Davie, K., Wouters, J., and Aerts, S. (2019). cisTopic: cis-regulatory
drive immunotherapy-related myocarditis. Nature 611, 818–826. topic modeling on single-cell ATAC-seq data. Nat Methods 16, 397–400.
Bagnoli, J.W., Ziegenhain, C., Janjic, A., Wange, L.E., Vieth, B., Parekh, S., Geuder, J., Brendel, M., Su, C., Bai, Z., Zhang, H., Elemento, O., and Wang, F. (2022). Application
Hellmann, I., and Enard, W. (2018). Sensitive and powerful single-cell RNA of deep learning on single-cell RNA sequencing data analysis: a review. Genomics
sequencing using mcSCRB-seq. Nat Commun 9, 2937. Proteomics Bioinf 20, 814–835.
Bai, Z., Deng, Y., Kim, D., Chen, Z., Xiao, Y., and Fan, R. (2020). An integrated Bressan, D., Battistoni, G., and Hannon, G.J. (2023). The dawn of spatial omics.
dielectrophoresis-trapping and nanowell transfer approach to enable double-sub- Science 381, eabq4964.
poisson single-cell RNA sequencing. ACS Nano 14, 7412–7424. Briggs, J.A., Weinreb, C., Wagner, D.E., Megason, S., Peshkin, L., Kirschner, M.W.,
Bannister, A.J., and Kouzarides, T. (2011). Regulation of chromatin by histone and Klein, A.M. (2018). The dynamics of gene expression in vertebrate
modifications. Cell Res 21, 381–395. embryogenesis at single-cell resolution. Science 360, eaar5780.
Barkas, N., Petukhov, V., Nikolaeva, D., Lozinsky, Y., Demharter, S., Khodosevich, K., Browaeys, R., Saelens, W., and Saeys, Y. (2020). NicheNet: modeling intercellular
and Kharchenko, P.V. (2019). Joint analysis of heterogeneous single-cell RNA-seq communication by linking ligands to target genes. Nat Methods 17, 159–162.
dataset collections. Nat Methods 16, 695–698. Buenrostro, J.D., Wu, B., Litzenburger, U.M., Ruff, D., Gonzales, M.L., Snyder, M.P.,
Barkley, D., Moncada, R., Pour, M., Liberman, D.A., Dryg, I., Werba, G., Wang, W., Chang, H.Y., and Greenleaf, W.J. (2015). Single-cell chromatin accessibility reveals
Baron, M., Rao, A., Xia, B., et al. (2022). Cancer cell states recur across tumor types principles of regulatory variation. Nature 523, 486–490.
and form specific interactions with the tumor microenvironment. Nat Genet 54, Butler, A., Hoffman, P., Smibert, P., Papalexi, E., and Satija, R. (2018). Integrating
1192–1201. single-cell transcriptomic data across different conditions, technologies, and species.
Baron, M., Veres, A., Wolock, S.L., Faust, A.L., Gaujoux, R., Vetere, A., Ryu, J.H., Nat Biotechnol 36, 411–420.
Wagner, B.K., Shen-Orr, S.S., Klein, A.M., et al. (2016). A single-cell transcriptomic Büttner, M., Miao, Z., Wolf, F.A., Teichmann, S.A., and Theis, F.J. (2019). A test
map of the human and mouse pancreas reveals inter- and intra-cell population metric for assessing single-cell RNA-seq batch correction. Nat Methods 16, 43–49.
structure. Cell Syst 3, 346–360.e4. Cabello-Aguilar, S., Alame, M., Kon-Sun-Tack, F., Fau, C., Lacroix, M., and Colinge, J.
Bartlett, D.A., Dileep, V., Handa, T., Ohkawa, Y., Kimura, H., Henikoff, S., and Gilbert, (2020). SingleCellSignalR: inference of intercellular networks from single-cell
D.M. (2021). High-throughput single-cell epigenomic profiling by targeted transcriptomics. Nucleic Acids Res 48, e55.
insertion of promoters (TIP-seq). J Cell Biol 220, e202103078. Cao, J., Cusanovich, D.A., Ramani, V., Aghamirzaie, D., Pliner, H.A., Hill, A.J., Daza,
Bartosovic, M., Kabbe, M., and Castelo-Branco, G. (2021). Single-cell CUT&Tag R.M., McFaline-Figueroa, J.L., Packer, J.S., Christiansen, L., et al. (2018). Joint
profiles histone modifications and transcription factors in complex tissues. Nat profiling of chromatin accessibility and gene expression in thousands of single cells.
Biotechnol 39, 825–835. Science 361, 1380–1385.
Bassez, A., Vos, H., Van Dyck, L., Floris, G., Arijs, I., Desmedt, C., Boeckx, B., Vanden Cao, J., O’Day, D.R., Pliner, H.A., Kingsley, P.D., Deng, M., Daza, R.M., Zager, M.A.,
Bempt, M., Nevelsteen, I., Lambein, K., et al. (2021). A single-cell map of Aldinger, K.A., Blecher-Gonen, R., Zhang, F., et al. (2020a). A human cell atlas of
intratumoral changes during anti-PD1 treatment of patients with breast cancer. fetal gene expression. Science 370, eaba7721.
Nat Med 27, 820–832. Cao, J., Zhou, W., Steemers, F., Trapnell, C., and Shendure, J. (2020b). Sci-fate
Bawa, G., Liu, Z., Yu, X., Qin, A., and Sun, X. (2022). Single-cell RNA sequencing for characterizes the dynamics of gene expression in single cells. Nat Biotechnol 38,
plant research: insights and possible benefits. Int J Mol Sci 23, 4497. 980–988.
Baysoy, A., Bai, Z., Satija, R., and Fan, R. (2023). The technological landscape and Cao, J., Packer, J.S., Ramani, V., Cusanovich, D.A., Huynh, C., Daza, R., Qiu, X., Lee,
applications of single-cell multi-omics. Nat Rev Mol Cell Biol 24, 695–713. C., Furlan, S.N., Steemers, F.J., et al. (2017). Comprehensive single-cell transcrip-
Bhattarai-Kline, S., Lear, S.K., Fishman, C.B., Lopez, S.C., Lockshin, E.R., Schubert, M. tional profiling of a multicellular organism. Science 357, 661–667.
G., Nivala, J., Church, G.M., and Shipman, S.L. (2022). Recording gene expression Cao, J., Spielmann, M., Qiu, X., Huang, X., Ibrahim, D.M., Hill, A.J., Zhang, F.,
order in DNA by CRISPR addition of retron barcodes. Nature 608, 217–225. Mundlos, S., Christiansen, L., Steemers, F.J., et al. (2019). The single-cell
Bian, H., Chen, Y., Dong, X., Li, C., Hao, M., Chen, S., Hu, J., Sun, M., Wei, L., and transcriptional landscape of mammalian organogenesis. Nature 566, 496–502.
Zhang, X. (2024). scMulan: a multitask generative pre-trained language model for Cao, Y., Bai, Y., Yuan, T., Song, L., Fan, Y., Ren, L., Song, W., Peng, J., An, R., Gu, Q.,
single-cell analysis. In: Ma, J., ed. Research in Computational Molecular Biology. et al. (2023). Single-cell bisulfite-free 5mC and 5hmC sequencing with high
RECOMB 2024. Lecture Notes in Computer Science. Cham: Springer. 479–482. sensitivity and scalability. Proc Natl Acad Sci USA 120, e2310367120.
Bian, S., Hou, Y., Zhou, X., Li, X., Yong, J., Wang, Y., Wang, W., Yan, J., Hu, B., Guo, Cao, Y., Su, B., Guo, X., Sun, W., Deng, Y., Bao, L., Zhu, Q., Zhang, X., Zheng, Y.,
H., et al. (2018). Single-cell multiomics sequencing and analyses of human Geng, C., et al. (2020c). Potent neutralizing antibodies against SARS-CoV-2
colorectal cancer. Science 362, 1060–1063. identified by high-throughput single-cell sequencing of convalescent patients’ B
Biancalani, T., Scalia, G., Buffoni, L., Avasthi, R., Lu, Z., Sanger, A., Tokcan, N., cells. Cell 182, 73–84.e16.
Vanderburg, C.R., Segerstolpe, Å., Zhang, M., et al. (2021). Deep learning and Cao, Z.J., and Gao, G. (2022). Multi-omics single-cell data integration and regulatory
alignment of spatially resolved single-cell transcriptomes with Tangram. Nat inference with graph-linked embedding. Nat Biotechnol 40, 1458–1466.
Methods 18, 1352–1362. Cardozo Gizzi, A.M., Cattoni, D.I., Fiche, J.B., Espinola, S.M., Gurgo, J., Messina, O.,
Bianchi, A., Scherer, M., Zaurin, R., Quililan, K., Velten, L., and Beekman, R. (2022). Houbron, C., Ogiyama, Y., Papadopoulos, G.L., Cavalli, G., et al. (2019).
scTAM-seq enables targeted high-confidence analysis of DNA methylation in single Microscopy-based chromosome conformation capture enables simultaneous
cells. Genome Biol 23, 229. visualization of genome organization and transcription in intact organisms. Mol
Blank, C.U., Haining, W.N., Held, W., Hogan, P.G., Kallies, A., Lugli, E., Lynn, R.C., Cell 74, 212–222.e5.
Philip, M., Rao, A., Restifo, N.P., et al. (2019). Defining ‘T cell exhaustion’. Nat Rev Carter, B., Ku, W.L., Kang, J.Y., Hu, G., Perrie, J., Tang, Q., and Zhao, K. (2019).
Immunol 19, 665–674. Mapping histone modifications in low cell number and single cells using antibody-
Blattman, S.B., Jiang, W., Oikonomou, P., and Tavazoie, S. (2020). Prokaryotic single- guided chromatin tagmentation (ACT-seq). Nat Commun 10, 3747.
cell RNA sequencing by in situ combinatorial indexing. Nat Microbiol 5, 1192– Cascone, T., Leung, C.H., Weissferdt, A., Pataer, A., Carter, B.W., Godoy, M.C.B.,
1201. Feldman, H., William Jr, W.N., Xi, Y., Basu, S., et al. (2023). Neoadjuvant
Bock, C., Datlinger, P., Chardon, F., Coelho, M.A., Dong, M.B., Lawson, K.A., Lu, T., chemotherapy plus nivolumab with or without ipilimumab in operable non-small
Maroc, L., Norman, T.M., Song, B., et al. (2022). High-content CRISPR screening. cell lung cancer: the phase 2 platform NEOSTAR trial. Nat Med 29, 593–604.
Nat Rev Methods Primers 2, 8. Chai, H., Tjong, H., Li, P., Liao, W., Wang, P., Wong, C.H., Ngan, C.Y., Leonard, W.J.,
Borgsmüller, N., Bonet, J., Marass, F., Gonzalez-Perez, A., Lopez-Bigas, N., Wei, C.L., and Ruan, Y. (2023). ChIATAC is an efficient strategy for multi-omics
Beerenwinkel, N., and Mathelier, A. (2020). BnpC: Bayesian non-parametric mapping of 3D epigenomes from low-cell inputs. Nat Commun 14, 213.
clustering of single-cell mutation profiles. Bioinformatics 36, 4854–4859. Chamorro González, R., Conrad, T., Stöber, M.C., Xu, R., Giurgiu, M., Rodriguez-Fos,
Borm, L.E., Mossi Albiach, A., Mannens, C.C.A., Janusauskas, J., Özgün, C., E., Kasack, K., Brückner, L., van Leen, E., Helmsauer, K., et al. (2023). Parallel
Fernández-García, D., Hodge, R., Castillo, F., Hedin, C.R.H., Villablanca, E.J., et sequencing of extrachromosomal circular DNAs and transcriptomes in single
al. (2023). Scalable in situ single-cell profiling by electrophoretic capture of mRNA cancer cells. Nat Genet 55, 880–890.

1268 SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 [Link]
Chang, L., Deng, E., Wang, J., Zhou, W., Ao, J., Liu, R., Su, D., and Fan, X. (2023a). 625–640.
Single-cell third-generation sequencing-based multi-omics uncovers gene expres- Cheng, S., Li, Z., Gao, R., Xing, B., Gao, Y., Yang, Y., Qin, S., Zhang, L., Ouyang, H.,
sion changes governed by ecDNA and structural variants in cancer cells. Clin Du, P., et al. (2021b). A pan-cancer single-cell transcriptional atlas of tumor
Transl Med 13, e1351. infiltrating myeloid cells. Cell 184, 792–809.e23.
Chang, L., Ruiz, P., Ito, T., and Sellers, W.R. (2021). Targeting pan-essential genes in Cheng, Y., Ma, X., Yuan, L., Sun, Z., and Wang, P. (2023b). Evaluating imputation
cancer: challenges and opportunities. Cancer Cell 39, 466–479. methods for single-cell RNA-seq data. BMC Bioinf 24, 302.
Chang, T., Han, W., Jiang, M., Li, J., Liao, Z., Tang, M., Zhang, J., Shen, J., Chen, Z., Chiou, J., Geusz, R.J., Okino, M.L., Han, J.Y., Miller, M., Melton, R., Beebe, E., Benaglio,
Fei, P., et al. (2023b). Rapid and signal crowdedness-robust in situ sequencing P., Huang, S., Korgaonkar, K., et al. (2021). Interpreting type 1 diabetes risk with
through hybrid block coding. Proc Natl Acad Sci USA 120, e2309227120. genetics and single-cell epigenomics. Nature 594, 398–402.
Charlton, J., Downing, T.L., Smith, Z.D., Gu, H., Clement, K., Pop, R., Akopian, V., Cho, C.S., Xi, J., Si, Y., Park, S.R., Hsu, J.E., Kim, M., Jun, G., Kang, H.M., and Lee, J.H.
Klages, S., Santos, D.P., Tsankov, A.M., et al. (2018). Author Correction: Global (2021). Microscopic examination of spatial transcriptome using Seq-Scope. Cell
delay in nascent strand DNA methylation. Nat Struct Mol Biol 25, 355. 184, 3559–3572.e22.
Chartrand, T., Dalley, R., Close, J., Goriounova, N.A., Lee, B.R., Mann, R., Miller, J.A., Chung, H., Melnikov, A., McCabe, C., Drokhlyansky, E., Wittenberghe, N.V., Magee, E.
Molnar, G., Mukora, A., Alfiler, L., et al. (2023). Morphoelectric and transcriptomic M., Waldman, J., Spira, A., Chen, F., Mazzilli, S., et al. (2022). SnFFPE-Seq: towards
divergence of the layer 1 interneuron repertoire in human versus mouse neocortex. scalable single nucleus RNA-Seq of formalin-fixed paraffin-embedded (FFPE) tissue.
Science 382, eadf0805. bioRxiv, 2022.2008.2025.505257.
Chen, A., Liao, S., Cheng, M., Ma, K., Wu, L., Lai, Y., Qiu, X., Yang, J., Xu, J., Hao, S., Chung, H., Parkhurst, C.N., Magee, E.M., Phillips, D., Habibi, E., Chen, F., Yeung, B.Z.,
et al. (2022a). Spatiotemporal transcriptomic atlas of mouse organogenesis using Waldman, J., Artis, D., and Regev, A. (2021). Joint single-cell measurements of
DNA nanoball-patterned arrays. Cell 185, 1777–1792.e21. nuclear proteins and RNA in vivo. Nat Methods 18, 1204–1212.
Chen, A., Sun, Y., Lei, Y., Li, C., Liao, S., Meng, J., Bai, Y., Liu, Z., Liang, Z., Zhu, Z., et Claringbould, A., and Zaugg, J.B. (2021). Enhancers in disease: molecular basis and
al. (2023a). Single-cell spatial transcriptome reveals cell-type organization in the emerging treatment strategies. Trends Mol Med 27, 1060–1073.
macaque cortex. Cell 186, 3726–3743.e24. Clark, I.C., Fontanez, K.M., Meltzer, R.H., Xue, Y., Hayford, C., May-Zhang, A.,
Chen, C., Xing, D., Tan, L., Li, H., Zhou, G., Huang, L., and Xie, X.S. (2017a). Single- D’Amato, C., Osman, A., Zhang, J.Q., Hettige, P., et al. (2023). Microfluidics-free
cell whole-genome analyses by Linear Amplification via Transposon Insertion single-cell genomics with templated emulsification. Nat Biotechnol 41, 1557–
(LIANTI). Science 356, 189–194. 1566.
Chen, H., Fang, X., Shao, J., Zhang, Q., Xu, L., Chen, J., Mei, Y., Jiang, M., Wang, Y., Clark, S.J., Argelaguet, R., Kapourani, C.A., Stubbs, T.M., Lee, H.J., Alda-Catalinas, C.,
Li, Z., et al. (2024a). Pan-cancer single-nucleus total RNA sequencing using snHH- Krueger, F., Sanguinetti, G., Kelsey, G., Marioni, J.C., et al. (2018). scNMT-seq
Seq. Adv Sci 11, e2304755. enables joint profiling of chromatin accessibility DNA methylation and transcrip-
Chen, H., Liao, Y., Zhang, G., Sun, Z., Yang, L., Fang, X., Sun, H., Ma, L., Fu, Y., Li, J., tion in single cells. Nat Commun 9, 781.
et al. (2021a). High-throughput Microwell-seq 2.0 profiles massively multiplexed Codeluppi, S., Borm, L.E., Zeisel, A., La Manno, G., van Lunteren, J.A., Svensson, C.I.,
chemical perturbation. Cell Discov 7, 107. and Linnarsson, S. (2018). Spatial organization of the somatosensory cortex
Chen, J., Suo, S., Tam, P.P., Han, J.D.J., Peng, G., and Jing, N. (2017b). Spatial revealed by osmFISH. Nat Methods 15, 932–935.
transcriptomic analysis of cryosectioned tissue samples with Geo-seq. Nat Protoc Crow, M., Paul, A., Ballouz, S., Huang, Z.J., and Gillis, J. (2018). Characterizing the
12, 566–580. replicability of cell types defined by single cell RNA-sequencing data using
Chen, J., Xu, H., Tao, W., Chen, Z., Zhao, Y., and Han, J.D.J. (2023b). Transformer for MetaNeighbor. Nat Commun 9, 884.
one stop interpretable cell type annotation. Nat Commun 14, 223. Cui, H., Wang, C., Maan, H., Pang, K., Luo, F., Duan, N., and Wang, B. (2024). scGPT:
Chen, K., and Wang, Z. (2023). A micropillar array based microfluidic device for rare toward building a foundation model for single-cell multi-omics using generative AI.
cell detection and single-cell proteomics. Methods Protoc 6, 80. Nat Methods 21, 1470–1480.
Chen, K.H., Boettiger, A.N., Moffitt, J.R., Wang, S., and Zhuang, X. (2015). Spatially Cuomo, A.S.E., Nathan, A., Raychaudhuri, S., MacArthur, D.G., and Powell, J.E.
resolved, highly multiplexed RNA profiling in single cells. Science 348, aaa6090. (2023). Single-cell genomics meets human genetics. Nat Rev Genet 24, 535–549.
Chen, L., and Flies, D.B. (2013). Molecular mechanisms of T cell co-stimulation and Cusanovich, D.A., Hill, A.J., Aghamirzaie, D., Daza, R.M., Pliner, H.A., Berletch, J.B.,
co-inhibition. Nat Rev Immunol 13, 227–242. Filippova, G.N., Huang, X., Christiansen, L., DeWitt, W.S., et al. (2018). A single-
Chen, S., Lake, B.B., and Zhang, K. (2019). High-throughput sequencing of the cell atlas of in vivo mammalian chromatin accessibility. Cell 174, 1309–1324.e18.
transcriptome and chromatin accessibility in the same cell. Nat Biotechnol 37, Dar, D., Dar, N., Cai, L., and Newman, D.K. (2021). Spatial transcriptomics of
1452–1457. planktonic and sessile bacterial populations at single-cell resolution. Science 373,
Chen, W., Guillaume-Gentil, O., Rainer, P.Y., Gäbelein, C.G., Saelens, W., Gardeux, V., eabi4882.
Klaeger, A., Dainese, R., Zachara, M., Zambelli, T., et al. (2022b). Live-seq enables Datlinger, P., Rendeiro, A.F., Boenke, T., Senekowitsch, M., Krausgruber, T., Barreca,
temporal transcriptomic recording of single cells. Nature 608, 733–740. D., and Bock, C. (2021). Ultra-high-throughput single-cell RNA sequencing and
Chen, W., Zhao, Y., Chen, X., Yang, Z., Xu, X., Bi, Y., Chen, V., Li, J., Choi, H., Ernest, perturbation screening with combinatorial fluidic indexing. Nat Methods 18, 635–
B., et al. (2021b). A multicenter study benchmarking single-cell RNA sequencing 642.
technologies using reference samples. Nat Biotechnol 39, 1103–1114. Datlinger, P., Rendeiro, A.F., Schmidl, C., Krausgruber, T., Traxler, P., Klughammer,
Chen, X., Miragaia, R.J., Natarajan, K.N., and Teichmann, S.A. (2018). A rapid and J., Schuster, L.C., Kuchler, A., Alpar, D., and Bock, C. (2017). Pooled CRISPR
robust method for single cell chromatin accessibility profiling. Nat Commun 9, screening with single-cell transcriptome readout. Nat Methods 14, 297–301.
5345. Davie, K., Janssens, J., Koldere, D., De Waegeneer, M., Pech, U., Kreft, Ł., Aibar, S.,
Chen, X., Wang, T., Chen, L., Zhao, Y., Deng, Y., Shen, W., Li, L., Yin, Z., Zhang, C., Makhzami, S., Christiaens, V., Bravo González-Blas, C., et al. (2018). A single-cell
Cai, G., et al. (2024b). Cross-species single-cell analysis uncovers the immuno- transcriptome atlas of the aging Drosophila brain. Cell 174, 982–998.e20.
pathological mechanisms associated with IgA nephropathy progression. JCI Insight de la Fuente, J., Legarra, N., Serrano, G., Osta, A.G., Kalari, K.R., Fernandez-Granda,
9, e173651. C., Ochoa, I., and Hernaez, M. (2023). Sweetwater: an interpretable and adaptive
Chen, Y., Wang, Y., Chen, Y., Cheng, Y., Wei, Y., Li, Y., Wang, J., Wei, Y., Chan, T.F., autoencoder for efficient tissue deconvolution. arXiv, 231111991.
and Li, Y. (2022c). Deep autoencoder for interpretable tissue-adaptive deconvolu- de Visser, K.E., and Joyce, J.A. (2023). The evolving tumor microenvironment: from
tion and cell-type-specific gene analysis. Nat Commun 13, 6735. cancer initiation to metastatic outgrowth. Cancer Cell 41, 374–403.
Chen, Z., Gong, F., Wan, L., and Ma, L. (2022d). BiTSC2: Bayesian inference of tumor Dekker, J., Belmont, A.S., Guttman, M., Leshyk, V.O., Lis, J.T., Lomvardas, S., Mirny, L.
clonal tree by joint analysis of single-cell SNV and CNA data. Brief Bioinf 23, A., O’Shea, C.C., Park, P.J., Ren, B., et al. (2017). The 4D nucleome project. Nature
bbac092. 549, 219–226.
Chen, Z., King, W.C., Hwang, A., Gerstein, M., and Zhang, J. (2022e). DeepVelo: single- Deng, C.C., Hu, Y.F., Zhu, D.H., Cheng, Q., Gu, J.J., Feng, Q.L., Zhang, L.X., Xu, Y.P.,
cell transcriptomic deep velocity field learning with neural ordinary differential Wang, D., Rong, Z., et al. (2021). Single-cell RNA-seq reveals fibroblast
equations. Sci Adv 8, eabq3745. heterogeneity and increased mesenchymal fibroblasts in human fibrotic skin
Cheng, H., Pui, H.P., Lentini, A., Kolbeinsdóttir, S., Andrews, N., Pei, Y., Reinius, B., diseases. Nat Commun 12, 3709.
Deng, Q., and Enge, M. (2021a). Smart3-ATAC: a highly sensitive method for joint Deng, Y., Bao, F., Dai, Q., Wu, L.F., and Altschuler, S.J. (2019). Scalable analysis of
accessibility and full-length transcriptome analysis in single cells. bioRxiv, cell-type composition from single-cell transcriptomics using deep recurrent
2021.2012.2002.470912. learning. Nat Methods 16, 311–314.
Cheng, M., Jiang, Y., Xu, J., Mentis, A.F.A., Wang, S., Zheng, H., Sahu, S.K., Liu, L., Deng, Y., Bartosovic, M., Kukanja, P., Zhang, D., Liu, Y., Su, G., Enninful, A., Bai, Z.,
and Xu, X. (2023a). Spatially resolved transcriptomics: a comprehensive review of Castelo-Branco, G., and Fan, R. (2022a). Spatial-CUT&Tag: spatially resolved
their technological advances, applications, and challenges. J Genet Genomics 50, chromatin modification profiling at the cellular level. Science 375, 681–686.

[Link] SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 1269
Deng, Y., Bartosovic, M., Ma, S., Zhang, D., Kukanja, P., Xiao, Y., Su, G., Liu, Y., Qin, Fang, L., Chen, Q., Wei, C.H., Lu, Z., and Wang, K. (2023). Bioformer: an efficient
X., Rosoklija, G.B., et al. (2022b). Spatial profiling of chromatin accessibility in transformer language model for biomedical text mining. arXiv, 2302.01588.
mouse and human tissues. Nature 609, 375–383. Fang, R., Preissl, S., Li, Y., Hou, X., Lucero, J., Wang, X., Motamedi, A., Shiau, A.K.,
Denisenko, E., de Kock, L., Tan, A., Beasley, A.B., Beilin, M., Jones, M.E., Hou, R., Zhou, X., Xie, F., et al. (2021). Comprehensive analysis of single cell ATAC-seq data
Muirí, D.Ó., Bilic, S., Mohan, G.R.K.A., et al. (2024). Spatial transcriptomics reveals with SnapATAC. Nat Commun 12, 1337.
discrete tumour microenvironments and autocrine loops within ovarian cancer Fang, R., Xia, C., Close, J.L., Zhang, M., He, J., Huang, Z., Halpern, A.R., Long, B.,
subclones. Nat Commun 15, 2860. Miller, J.A., Lein, E.S., et al. (2022a). Conservation and divergence of cortical cell
Dey, S.S., Kester, L., Spanjaard, B., Bienko, M., and van Oudenaarden, A. (2015). organization in human and mouse revealed by MERFISH. Science 377, 56–62.
Integrated genome and transcriptome sequencing of the same cell. Nat Biotechnol Fang, X., Jiang, M., Zhou, M., Shao, J., Fang, X., Wang, J., Fu, Y., Liao, Y., Chen, H.,
33, 285–289. Wang, R., et al. (2022b). Elucidating the developmental dynamics of mouse
Dixit, A., Parnas, O., Li, B., Chen, J., Fulco, C.P., Jerby-Arnon, L., Marjanovic, N.D., stromal cells at single-cell level. Life Med 1, 45–48.
Dionne, D., Burks, T., Raychowdhury, R., et al. (2016). Perturb-Seq: dissecting Fawkner-Corbett, D., Antanaviciute, A., Parikh, K., Jagielowicz, M., Gerós, A.S.,
molecular circuits with scalable single-cell RNA profiling of pooled genetic screens. Gupta, T., Ashley, N., Khamis, D., Fowler, D., Morrissey, E., et al. (2021).
Cell 167, 1853–1866.e17. Spatiotemporal analysis of human intestinal development at single-cell resolution.
Domcke, S., Hill, A.J., Daza, R.M., Cao, J., O’Day, D.R., Pliner, H.A., Aldinger, K.A., Cell 184, 810–826.e23.
Pokholok, D., Zhang, F., Milbank, J.H., et al. (2020). A human cell atlas of fetal Fei, L., Chen, H., Ma, L., E, W., Wang, R., Fang, X., Zhou, Z., Sun, H., Wang, J., Jiang,
chromatin accessibility. Science 370, 809. M., et al. (2022). Systematic identification of cell-fate regulatory programs using a
Domínguez Conde, C., Xu, C., Jarvis, L.B., Rainbow, D.B., Wells, S.B., Gomes, T., single-cell atlas of mouse development. Nat Genet 54, 1051–1061.
Howlett, S.K., Suchanek, O., Polanski, K., King, H.W., et al. (2022). Cross-tissue Femino, A.M., Fay, F.S., Fogarty, K., and Singer, R.H. (1998). Visualization of single
immune cell analysis reveals tissue-specific features in humans. Science 376, RNA transcripts in situ. Science 280, 585–590.
eabl5197. Fiskin, E., Lareau, C.A., Ludwig, L.S., Eraslan, G., Liu, F., Ring, A.M., Xavier, R.J., and
Dong, K., and Zhang, S. (2022). Deciphering spatial domains from spatially resolved Regev, A. (2022). Single-cell profiling of proteins and chromatin accessibility using
transcriptomics with an adaptive graph attention auto-encoder. Nat Commun 13, PHAGE-ATAC. Nat Biotechnol 40, 374–381.
1739. Flyamer, I.M., Gassler, J., Imakaev, M., Brandão, H.B., Ulianov, S.V., Abdennur, N.,
Dong, M., Wang, B., Wei, J., de O. Fonseca, A.H., Perry, C.J., Frey, A., Ouerghi, F., Razin, S.V., Mirny, L.A., and Tachibana-Konwalski, K. (2017). Single-nucleus Hi-C
Foxman, E.F., Ishizuka, J.J., Dhodapkar, R.M., et al. (2023). Causal identification of reveals unique chromatin reorganization at oocyte-to-zygote transition. Nature
single-cell experimental perturbation effects with CINEMA-OT. Nat Methods 20, 544, 110–114.
1769–1779. Frangoul, H., Altshuler, D., Cappellini, M.D., Chen, Y.S., Domm, J., Eustace, B.K., Foell,
Du, J., Wang, Y., Chen, W., Xu, M., Zhou, R., Shou, H., and Chen, J. (2023). High- J., de la Fuente, J., Grupp, S., Handgretinger, R., et al. (2021). CRISPR-Cas9 gene
resolution anatomical and spatial transcriptome analyses reveal two types of editing for sickle cell disease and β-thalassemia. N Engl J Med 384, 252–260.
meristematic cell pools within the secondary vascular tissue of poplar stem. Mol Frank, M.H., Wilson, B.J., Gold, J.S., and Frank, N.Y. (2021). Clinical implications of
Plant 16, 809–828. colorectal cancer stem cells in the age of single-cell omics and targeted therapies.
Efremova, M., Vento-Tormo, M., Teichmann, S.A., and Vento-Tormo, R. (2020). Gastroenterology 160, 1947–1960.
CellPhoneDB: inferring cell-cell communication from combined expression of multi- Fu, X., Sun, L., Dong, R., Chen, J.Y., Silakit, R., Condon, L.F., Lin, Y., Lin, S., Palmiter,
subunit ligand-receptor complexes. Nat Protoc 15, 1484–1506. R.D., and Gu, L. (2022). Polony gels enable amplifiable DNA stamping and spatial
Eisenstein, M. (2020). Single-cell RNA-seq analysis software providers scramble to transcriptomics of chronic pain. Cell 185, 4621–4633.e17.
offer solutions. Nat Biotechnol 38, 254–257. Fullwood, M.J., Liu, M.H., Pan, Y.F., Liu, J., Xu, H., Mohamed, Y.B., Orlov, Y.L.,
Elmentaite, R., Domínguez Conde, C., Yang, L., and Teichmann, S.A. (2022). Single- Velkov, S., Ho, A., Mei, P.H., et al. (2009). An oestrogen-receptor-α-bound human
cell atlases: Shared and tissue-specific cell types across human organs. Nat Rev chromatin interactome. Nature 462, 58–64.
Genet 23, 395–410. Funnell, T., O’Flanagan, C.H., Williams, M.J., McPherson, A., McKinney, S., Kabeer,
Emmert-Buck, M.R., Bonner, R.F., Smith, P.D., Chuaqui, R.F., Zhuang, Z., Goldstein, F., Lee, H., Salehi, S., Vázquez-García, I., Shi, H., et al. (2022). Single-cell genomic
S.R., Weiss, R.A., and Liotta, L.A. (1996). Laser capture microdissection. Science variation induced by mutational processes in cancer. Nature 612, 106–115.
274, 998–1001. Galluzzi, L., Buqué, A., Kepp, O., Zitvogel, L., and Kroemer, G. (2015). Immunological
Eng, C.H.L., Lawson, M., Zhu, Q., Dries, R., Koulena, N., Takei, Y., Yun, J., Cronin, C., effects of conventional chemotherapy and targeted anticancer agents. Cancer Cell
Karp, C., Yuan, G.C., et al. (2019). Transcriptome-scale super-resolved imaging in 28, 690–714.
tissues by RNA seqFISH+. Nature 568, 235–239. Gangadhar, T.C., and Vonderheide, R.H. (2014). Mitigating the toxic effects of
Engblom, C., Pfirschke, C., and Pittet, M.J. (2016). The role of myeloid cells in cancer anticancer immunotherapy. Nat Rev Clin Oncol 11, 91–99.
therapies. Nat Rev Cancer 16, 447–462. Gao, W., Ku, W.L., Pan, L., Perrie, J., Zhao, T., Hu, G., Wu, Y., Zhu, J., Ni, B., and
Eraslan, G., Drokhlyansky, E., Anand, S., Fiskin, E., Subramanian, A., Slyper, M., Zhao, K. (2021). Multiplex indexing approach for the detection of DNase I
Wang, J., Van Wittenberghe, N., Rouhana, J.M., Waldman, J., et al. (2022). Single- hypersensitive sites in single cells. Nucleic Acids Res 49, e56.
nucleus cross-tissue molecular reference maps toward understanding disease gene Gao, Y., Yi, L., Zhan, J., Wang, L., Yao, X., Yan, J., Jian, S., Gao, L., Farangez, M., Gao,
function. Science 376, eabl4290. M., et al. (2023). A clinical study of preimplantation DNA methylation screening in
Eraslan, G., Simon, L.M., Mircea, M., Mueller, N.S., and Theis, F.J. (2019). Single-cell assisted reproductive technology. Cell Res 33, 483–485.
RNA-seq denoising using a deep count autoencoder. Nat Commun 10, 390. Garcia, F.J., Sun, N., Lee, H., Godlewski, B., Mathys, H., Galani, K., Zhou, B., Jiang, X.,
Erfanian, N., Heydari, A.A., Feriz, A.M., Iañez, P., Derakhshani, A., Ghasemigol, M., Ng, A.P., Mantero, J., et al. (2022). Single-cell dissection of the human brain
Farahpour, M., Razavi, S.M., Nasseri, S., Safarpour, H., et al. (2023). Deep learning vasculature. Nature 603, 893–899.
applications in single-cell genomics and transcriptomics data analysis. Biomed Gavish, A., Tyler, M., Greenwald, A.C., Hoefflin, R., Simkin, D., Tschernichovsky, R.,
Pharmacother 165, 115077. Galili Darnell, N., Somech, E., Barbolin, C., Antman, T., et al. (2023). Hallmarks of
Erhard, F., Baptista, M.A.P., Krammer, T., Hennig, T., Lange, M., Arampatzi, P., transcriptional intratumour heterogeneity across a thousand tumours. Nature 618,
Jürges, C.S., Theis, F.J., Saliba, A.E., and Dölken, L. (2019). scSLAM-seq reveals 598–606.
core features of transcription dynamics in single cells. Nature 571, 419–423. Gehring, J., Hwee Park, J., Chen, S., Thomson, M., and Pachter, L. (2020). Highly
Erickson, A., He, M., Berglund, E., Marklund, M., Mirzazadeh, R., Schultz, N., Kvastad, multiplexed single-cell RNA-seq by DNA oligonucleotide tagging of cellular
L., Andersson, A., Bergenstråhle, L., Bergenstråhle, J., et al. (2022). Spatially proteins. Nat Biotechnol 38, 35–38.
resolved clonal copy number alterations in benign and malignant tissue. Nature Gerber, T., Murawala, P., Knapp, D., Masselink, W., Schuez, M., Hermann, S., Gac-
608, 360–367. Santel, M., Nowoshilow, S., Kageyama, J., Khattak, S., et al. (2018). Single-cell
Falcão, A.M., van Bruggen, D., Marques, S., Meijer, M., Jäkel, S., Agirre, E., analysis uncovers convergence of cell identities during axolotl limb regeneration.
Samudyata, E., Floriddia, E.M., Vanichkina, D.P., ffrench-Constant, C., et al. Science 362, eaaq0681.
(2018). Disease-specific oligodendrocyte lineage cells arise in multiple sclerosis. Nat Gerlach, J.P., van Buggenum, J.A.G., Tanis, S.E.J., Hogeweg, M., Heuts, B.M.H.,
Med 24, 1837–1844. Muraro, M.J., Elze, L., Rivello, F., Rakszewska, A., van Oudenaarden, A., et al.
Fan, H.C., Fu, G.K., and Fodor, S.P.A. (2015a). Combinatorial labeling of single cells (2019). Combined quantification of intracellular (phospho-)proteins and transcrip-
for gene expression cytometry. Science 347, 1258367. tomics from fixed single cells. Sci Rep 9, 1469.
Fan, X., Zhang, X., Wu, X., Guo, H., Hu, Y., Tang, F., and Huang, Y. (2015b). Single- Gerlinger, M., Rowan, A.J., Horswell, S., Larkin, J., Endesfelder, D., Gronroos, E.,
cell RNA-seq transcriptome analysis of linear and circular RNAs in mouse Martinez, P., Matthews, N., Stewart, A., Tarpey, P., et al. (2012). Intratumor
preimplantation embryos. Genome Biol 16, 148. heterogeneity and branched evolution revealed by multiregion sequencing. N Engl J

1270 SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 [Link]
 Med 366, 883–892. intestinal epithelium. Nature 551, 333–339.
Gerstberger, S., Jiang, Q., and Ganesh, K. (2023). Metastasis. Cell 186, 1564–1579. Hagemann-Jensen, M., Ziegenhain, C., and Sandberg, R. (2022). Scalable single-cell
Ghazanfar, S., Guibentif, C., and Marioni, J.C. (2024). Stabilized mosaic single-cell RNA sequencing from full transcripts with Smart-seq3xpress. Nat Biotechnol 40,
data integration using unshared features. Nat Biotechnol 42, 284–292. 1452–1457.
Giannoukos, G., Ciulla, D.M., Huang, K., Haas, B.J., Izard, J., Levin, J.Z., Livny, J., Earl, Haghverdi, L., Büttner, M., Wolf, F.A., Buettner, F., and Theis, F.J. (2016). Diffusion
A.M., Gevers, D., Ward, D.V., et al. (2012). Efficient and robust RNA-seq process for pseudotime robustly reconstructs lineage branching. Nat Methods 13, 845–848.
cultured bacteria and complex community transcriptomes. Genome Biol 13, R23. Haghverdi, L., Lun, A.T.L., Morgan, M.D., and Marioni, J.C. (2018). Batch effects in
Gierahn, T.M., Wadsworth Ii, M.H., Hughes, T.K., Bryson, B.D., Butler, A., Satija, R., single-cell RNA-sequencing data are corrected by matching mutual nearest
Fortune, S., Love, J.C., and Shalek, A.K. (2017). Erratum: Seq-Well: portable, low- neighbors. Nat Biotechnol 36, 421–427.
cost RNA sequencing of single cells at high throughput. Nat Methods 14, 752. Hainer, S.J., Bošković, A., McCannell, K.N., Rando, O.J., and Fazzio, T.G. (2019).
Giolai, M., Verweij, W., Lister, A., Heavens, D., Macaulay, I., and Clark, M.D. (2019). Profiling of pluripotency factors in single cells and early embryos. Cell 177, 1319–
Spatially resolved transcriptomics reveals plant host responses to pathogens. Plant 1329.e11.
Methods 15, 114. Hamashima, K., Wong, K.W., Sam, T.W., Teo, J.H.J., Taneja, R., Le, M.T.N., Li, Q.J.,
Goldstein, L.D., Chen, Y.J.J., Wu, J., Chaudhuri, S., Hsiao, Y.C., Schneider, K., Hoi, K. Hanna, J.H., Li, H., and Loh, Y.H. (2023). Single-nucleus multiomic mapping of
H., Lin, Z., Guerrero, S., Jaiswal, B.S., et al. (2019). Massively parallel single-cell B- m6A methylomes and transcriptomes in native populations of cells with sn-m6A-
cell receptor sequencing enables rapid discovery of diverse antigen-reactive CT. Mol Cell 83, 3205–3216.e5.
antibodies. Commun Biol 2, 304. Han, K.Y., Kim, K.T., Joung, J.G., Son, D.S., Kim, Y.J., Jo, A., Jeon, H.J., Moon, H.S.,
Gong, B., Zhou, Y., and Purdom, E. (2021). Cobolt: Integrative analysis of multimodal Yoo, C.E., Chung, W., et al. (2018a). SIDR: simultaneous isolation and parallel
single-cell sequencing data. Genome Biol 22, 351. sequencing of genomic DNA and total RNA from single cells. Genome Res 28, 75–
Gong, W., Kwak, I.Y., Pota, P., Koyano-Nakagawa, N., and Garry, D.J. (2018). 87.
DrImpute: Imputing dropout events in single cell RNA sequencing data. BMC Bioinf Han, L., Wei, X., Liu, C., Volpe, G., Zhuang, Z., Zou, X., Wang, Z., Pan, T., Yuan, Y.,
19, 220. Zhang, X., et al. (2022). Cell transcriptomic atlas of the non-human primate
Gopalan, S., Wang, Y., Harper, N.W., Garber, M., and Fazzio, T.G. (2021). Macaca fascicularis. Nature 604, 723–731.
Simultaneous profiling of multiple chromatin proteins in the same cells. Mol Cell Han, L., Zi, X., Garmire, L.X., Wu, Y., Weissman, S.M., Pan, X., and Fan, R. (2014).
81, 4736–4746.e5. Co-detection and sequencing of genes and transcripts from the same single cells
Gorkin, D.U., Barozzi, I., Zhao, Y., Zhang, Y., Huang, H., Lee, A.Y., Li, B., Chiou, J., facilitated by a microfluidics platform. Sci Rep 4, 6485.
Wildberg, A., Ding, B., et al. (2020). An atlas of dynamic chromatin landscapes in Han, X., Wang, R., Zhou, Y., Fei, L., Sun, H., Lai, S., Saadatpour, A., Zhou, Z., Chen,
mouse fetal development. Nature 583, 744–751. H., Ye, F., et al. (2018b). Mapping the mouse cell atlas by microwell-seq. Cell 173,
Granja, J.M., Corces, M.R., Pierce, S.E., Bagdatli, S.T., Choudhry, H., Chang, H.Y., and 1307.
Greenleaf, W.J. (2021). Author Correction: ArchR is a scalable software package for Han, X., Zhou, Z., Fei, L., Sun, H., Wang, R., Chen, Y., Chen, H., Wang, J., Tang, H.,
integrative single-cell chromatin accessibility analysis. Nat Genet 53, 935. Ge, W., et al. (2020). Construction of a human cell landscape at single-cell level.
Greenbaum, S., Averbukh, I., Soon, E., Rizzuto, G., Baranski, A., Greenwald, N.F., Nature 581, 303–309.
Kagel, A., Bosse, M., Jaswa, E.G., Khair, Z., et al. (2023). A spatially resolved Hanahan, D. (2022). Hallmarks of cancer: new dimensions. Cancer Discov 12, 31–46.
timeline of the human maternal-fetal interface. Nature 619, 595–605. Haniffa, M., Taylor, D., Linnarsson, S., Aronow, B.J., Bader, G.D., Barker, R.A.,
Greenman, C., Stephens, P., Smith, R., Dalgliesh, G.L., Hunter, C., Bignell, G., Davies, Camara, P.G., Camp, J.G., Chédotal, A., Copp, A., et al. (2021). A roadmap for the
H., Teague, J., Butler, A., Stevens, C., et al. (2007). Patterns of somatic mutation in Human Developmental Cell Atlas. Nature 597, 196–205.
human cancer genomes. Nature 446, 153–158. Hao, M., Gong, J., Zeng, X., Liu, C., Guo, Y., Cheng, X., Wang, T., Ma, J., Zhang, X.,
Gross, A., Schoendube, J., Zimmermann, S., Steeb, M., Zengerle, R., and Koltay, P. and Song, L. (2024a). Large-scale foundation model on single-cell transcriptomics.
(2015). Technologies for single-cell isolation. Int J Mol Sci 16, 16897–16919. Nat Methods 21, 1481–1491.
Grosselin, K., Durand, A., Marsolier, J., Poitou, A., Marangoni, E., Nemati, F., Hao, Y., Hao, S., Andersen-Nissen, E., Mauck Iii, W.M., Zheng, S., Butler, A., Lee, M.J.,
Dahmani, A., Lameiras, S., Reyal, F., Frenoy, O., et al. (2019). High-throughput Wilk, A.J., Darby, C., Zager, M., et al. (2021). Integrated analysis of multimodal
single-cell ChIP-seq identifies heterogeneity of chromatin states in breast cancer. single-cell data. Cell 184, 3573–3587.e29.
Nat Genet 51, 1060–1066. Hao, Y., Stuart, T., Kowalski, M.H., Choudhary, S., Hoffman, P., Hartman, A.,
Gu, H., Raman, A.T., Wang, X., Gaiti, F., Chaligne, R., Mohammad, A.W., Arczewska, Srivastava, A., Molla, G., Madad, S., Fernandez-Granda, C., et al. (2024b).
A., Smith, Z.D., Landau, D.A., Aryee, M.J., et al. (2021). Smart-RRBS for single-cell Dictionary learning for integrative, multimodal and scalable single-cell analysis.
methylome and transcriptome analysis. Nat Protoc 16, 4004–4030. Nat Biotechnol 42, 293–304.
Gu, W., Crawford, E.D., O’Donovan, B.D., Wilson, M.R., Chow, E.D., Retallack, H., and Hao, Z.Z., Wei, J.R., Xiao, D., Liu, R., Xu, N., Tang, L., Huang, M., Shen, Y., Xing, C.,
DeRisi, J.L. (2016). Depletion of Abundant Sequences by Hybridization (DASH): Huang, W., et al. (2022). Single-cell transcriptomics of adult macaque
using Cas9 to remove unwanted high-abundance species in sequencing libraries hippocampus reveals neural precursor cell populations. Nat Neurosci 25, 805–817.
and molecular counting applications. Genome Biol 17, 41. Harroud, A., and Hafler, D.A. (2023). Common genetic factors among autoimmune
Guinney, J., Dienstmann, R., Wang, X., de Reyniès, A., Schlicker, A., Soneson, C., diseases. Science 380, 485–490.
Marisa, L., Roepman, P., Nyamundanda, G., Angelino, P., et al. (2015). The Hartmann, F.J., Mrdjen, D., McCaffrey, E., Glass, D.R., Greenwald, N.F., Bharadwaj,
consensus molecular subtypes of colorectal cancer. Nat Med 21, 1350–1356. A., Khair, Z., Verberk, S.G.S., Baranski, A., Baskar, R., et al. (2021). Single-cell
Guo, C., Kong, W., Kamimoto, K., Rivera-Gonzalez, G.C., Yang, X., Kirita, Y., and metabolic profiling of human cytotoxic T cells. Nat Biotechnol 39, 186–197.
Morris, S.A. (2019). CellTag Indexing: genetic barcode-based sample multiplexing Hashimshony, T., Senderovich, N., Avital, G., Klochendler, A., de Leeuw, Y., Anavy,
for single-cell genomics. Genome Biol 20, 90. L., Gennert, D., Li, S., Livak, K.J., Rozenblatt-Rosen, O., et al. (2016). CEL-Seq2:
Guo, F., Li, L., Li, J., Wu, X., Hu, B., Zhu, P., Wen, L., and Tang, F. (2017). Single-cell Sensitive highly-multiplexed single-cell RNA-Seq. Genome Biol 17, 77.
multi-omics sequencing of mouse early embryos and embryonic stem cells. Cell Res Hashimshony, T., Wagner, F., Sher, N., and Yanai, I. (2012). CEL-Seq: single-cell
27, 967–988. RNA-Seq by multiplexed linear amplification. Cell Rep 2, 666–673.
Guo, H., Zhu, P., Wu, X., Li, X., Wen, L., and Tang, F. (2013). Single-cell methylome Hayashi, T., Ozaki, H., Sasagawa, Y., Umeda, M., Danno, H., and Nikaido, I. (2018).
landscapes of mouse embryonic stem cells and early embryos analyzed using Single-cell full-length total RNA sequencing uncovers dynamics of recursive
reduced representation bisulfite sequencing. Genome Res 23, 2126–2135. splicing and enhancer RNAs. Nat Commun 9, 619.
Guo, X., Xu, W., Zhang, W., Pan, C., Thalacker-Mercer, A.E., Zheng, H., and Gu, Z. He, H., Chen, S., Fan, Z., Dong, Y., Wang, Y., Li, S., Sun, X., Song, Y., Yang, J., Cao, Q.,
(2023). High-frequency and functional mitochondrial DNA mutations at the et al. (2023). Multi-dimensional single-cell characterization revealed suppressive
single-cell level. Proc Natl Acad Sci USA 120, e2201518120. immune microenvironment in AFP-positive hepatocellular carcinoma. Cell Discov
Guo, X., Zhang, Y., Zheng, L., Zheng, C., Song, J., Zhang, Q., Kang, B., Liu, Z., Jin, L., 9, 60.
Xing, R., et al. (2018). Global characterization of T cells in non-small-cell lung He, K., Zhang, X., Ren, S., and Sun, J. (2016). Deep residual learning for image
cancer by single-cell sequencing. Nat Med 24, 978–985. recognition. In: Proceedings of the IEEE Conference on Computer Vision and
Gyllborg, D., Langseth, C.M., Qian, X., Choi, E., Salas, S.M., Hilscher, M.M., Lein, E.S., Pattern Recognition (CVPR). Las Vegas. 770–778.
and Nilsson, M. (2020). Hybridization-based in situ sequencing (HybISS) for He, P., Lim, K., Sun, D., Pett, J.P., Jeng, Q., Polanski, K., Dong, Z., Bolt, L., Richardson,
spatially resolved transcriptomics in human and mouse brain tissue. Nucleic Acids L., Mamanova, L., et al. (2022). A human fetal lung cell atlas uncovers proximal-
Res 48, e112. distal gradients of differentiation and key regulators of epithelial fates. Cell 185,
Haber, A.L., Biton, M., Rogel, N., Herbst, R.H., Shekhar, K., Smillie, C., Burgin, G., 4841–4860.e25.
Delorey, T.M., Howitt, M.R., Katz, Y., et al. (2017). A single-cell survey of the small He, S., Wang, L.H., Liu, Y., Li, Y.Q., Chen, H.T., Xu, J.H., Peng, W., Lin, G.W., Wei, P.

[Link] SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 1271
 P., Li, B., et al. (2020). Single-cell transcriptome profiling of an adult human cell International Human Genome Sequencing Consortium (2004). Finishing the
atlas of 15 major organs. Genome Biol 21, 294. euchromatic sequence of the human genome. Nature 431, 931–945.
Hendriks, G.J., Jung, L.A., Larsson, A.J.M., Lidschreiber, M., Andersson Forsman, O., Islam, S., Kjällquist, U., Moliner, A., Zajac, P., Fan, J.B., Lönnerberg, P., and
Lidschreiber, K., Cramer, P., and Sandberg, R. (2019). NASC-seq monitors RNA Linnarsson, S. (2011). Characterization of the single-cell transcriptional landscape
synthesis in single cells. Nat Commun 10, 3138. by highly multiplex RNA-seq. Genome Res 21, 1160–1167.
Hickey, J.W., Becker, W.R., Nevins, S.A., Horning, A., Perez, A.E., Zhu, C., Zhu, B., Jackson, H.W., Fischer, J.R., Zanotelli, V.R.T., Ali, H.R., Mechera, R., Soysal, S.D.,
Wei, B., Chiu, R., Chen, D.C., et al. (2023). Organization of the human intestine at Moch, H., Muenst, S., Varga, Z., Weber, W.P., et al. (2020). The single-cell
single-cell resolution. Nature 619, 572–584. pathology landscape of breast cancer. Nature 578, 615–620.
Hie, B., Bryson, B., and Berger, B. (2019). Efficient integration of heterogeneous Jagadeesh, K.A., Dey, K.K., Montoro, D.T., Mohan, R., Gazal, S., Engreitz, J.M., Xavier,
single-cell transcriptomes using Scanorama. Nat Biotechnol 37, 685–691. R.J., Price, A.L., and Regev, A. (2022). Identifying disease-critical cell types and
Hill, A.V.S. (1998). The immunogenetics of human infectious diseases. Annu Rev cellular processes by integrating single-cell RNA-sequencing and human genetics.
Immunol 16, 593–617. Nat Genet 54, 1479–1492.
Hoadley, K.A., Yau, C., Hinoue, T., Wolf, D.M., Lazar, A.J., Drill, E., Shen, R., Taylor, Jaitin, D.A., Kenigsberg, E., Keren-Shaul, H., Elefant, N., Paul, F., Zaretsky, I., Mildner,
A.M., Cherniack, A.D., Thorsson, V., et al. (2018). Cell-of-origin patterns dominate A., Cohen, N., Jung, S., Tanay, A., et al. (2014). Massively parallel single-cell RNA-
the molecular classification of 10,000 tumors from 33 types of cancer. Cell 173, seq for marker-free decomposition of tissues into cell types. Science 343, 776–779.
291–304.e6. Jaitin, D.A., Weiner, A., Yofe, I., Lara-Astiaso, D., Keren-Shaul, H., David, E., Salame,
Hochgerner, H., Lönnerberg, P., Hodge, R., Mikes, J., Heskol, A., Hubschle, H., Lin, P., T.M., Tanay, A., van Oudenaarden, A., and Amit, I. (2016). Dissecting immune
Picelli, S., La Manno, G., Ratz, M., et al. (2017). STRT-seq-2i: dual-index 5ʹ single circuits by linking CRISPR-pooled screens with single-cell RNA-seq. Cell 167,
cell and nucleus RNA-seq on an addressable microwell array. Sci Rep 7, 16327. 1883–1896.e15.
Homberger, C., Hayward, R.J., Barquist, L., Vogel, J., and Laub, M.T. (2023). Janssens, D.H., Otto, D.J., Meers, M.P., Setty, M., Ahmad, K., and Henikoff, S. (2022a).
Improved bacterial single-cell RNA-seq through automated MATQ-seq and Cas9- CUT&Tag2for1: a modified method for simultaneous profiling of the accessible and
based removal of rRNA reads. mBio 14, e0355722. silenced regulome in single cells. Genome Biol 23, 81.
Honda, M., Oki, S., Kimura, R., Harada, A., Maehara, K., Tanaka, K., Meno, C., and Janssens, J., Aibar, S., Taskiran, I.I., Ismail, J.N., Gomez, A.E., Aughey, G., Spanier, K.
Ohkawa, Y. (2021). High-depth spatial transcriptome analysis by photo-isolation I., De Rop, F.V., González-Blas, C.B., Dionne, M., et al. (2022b). Decoding gene
chemistry. Nat Commun 12, 4416. regulation in the fly brain. Nature 601, 630–636.
Hou, W., and Ji, Z. (2024). Assessing GPT-4 for cell type annotation in single-cell Ji, A.L., Rubin, A.J., Thrane, K., Jiang, S., Reynolds, D.L., Meyers, R.M., Guo, M.G.,
RNA-seq analysis. Nat Methods 21, 1462–1465. George, B.M., Mollbrink, A., Bergenstråhle, J., et al. (2020). Multimodal analysis of
Hou, W., Ji, Z., Ji, H., and Hicks, S.C. (2020). A systematic evaluation of single-cell composition and spatial architecture in human squamous cell carcinoma. Cell 182,
RNA-sequencing imputation methods. Genome Biol 21, 218. 497–514.e22.
Hou, Y., Guo, H., Cao, C., Li, X., Hu, B., Zhu, P., Wu, X., Wen, L., Tang, F., Huang, Y., Jiang, F., Zhou, X., Qian, Y., Zhu, M., Wang, L., Li, Z., Shen, Q., Wang, M., Qu, F., Cui,
et al. (2016). Single-cell triple omics sequencing reveals genetic, epigenetic, and G., et al. (2023a). Simultaneous profiling of spatial gene expression and chromatin
transcriptomic heterogeneity in hepatocellular carcinomas. Cell Res 26, 304–319. accessibility during mouse brain development. Nat Methods 20, 1048–1057.
Hrovatin, K., Bastidas-Ponce, A., Bakhti, M., Zappia, L., Büttner, M., Salinno, C., Sterr, Jiang, M., Xiao, Y., E, W., Ma, L., Wang, J., Chen, H., Gao, C., Liao, Y., Guo, Q., Peng,
M., Böttcher, A., Migliorini, A., Lickert, H., et al. (2023). Delineating mouse β-cell J., et al. (2021). Characterization of the zebrafish cell landscape at single-cell
identity during lifetime and in diabetes with a single cell atlas. Nat Metab 5, 1615– resolution. Front Cell Dev Biol 9, 743421.
1637. Jiang, S., Huang, Z., Li, Y., Yu, C., Yu, H., Ke, Y., Jiang, L., and Liu, J. (2023b). Single-
Hu, J., Li, X., Coleman, K., Schroeder, A., Ma, N., Irwin, D.J., Lee, E.B., Shinohara, R. cell chromatin accessibility and transcriptome atlas of mouse embryos. Cell Rep 42,
T., and Li, M. (2021). SpaGCN: integrating gene expression, spatial location and 112210.
histology to identify spatial domains and spatially variable genes by graph Jiang, Y.R., Zhu, L., Cao, L.R., Wu, Q., Chen, J.B., Wang, Y., Wu, J., Zhang, T.Y.,
convolutional network. Nat Methods 18, 1342–1351. Wang, Z.L., Guan, Z.Y., et al. (2023c). Simultaneous deep transcriptome and
Hu, K.H., Eichorst, J.P., McGinnis, C.S., Patterson, D.M., Chow, E.D., Kersten, K., proteome profiling in a single mouse oocyte. Cell Rep 42, 113455.
Jameson, S.C., Gartner, Z.J., Rao, A.A., and Krummel, M.F. (2020). ZipSeq: Jiao, L., Wang, G., Dai, H., Li, X., Wang, S., and Song, T. (2023). scTransSort:
barcoding for real-time mapping of single cell transcriptomes. Nat Methods 17, transformers for intelligent annotation of cell types by gene embeddings.
833–843. Biomolecules 13, 611.
Hu, P., Zhang, W., Xin, H., and Deng, G. (2016a). Single cell isolation and analysis. Jin, S., Guerrero-Juarez, C.F., Zhang, L., Chang, I., Ramos, R., Kuan, C.H., Myung, P.,
Front Cell Dev Biol 4, 116. Plikus, M.V., and Nie, Q. (2021). Inference and analysis of cell-cell communication
Hu, T., Allam, M., Cai, S., Henderson, W., Yueh, B., Garipcan, A., Ievlev, A.V., using CellChat. Nat Commun 12, 1088.
Afkarian, M., Beyaz, S., and Coskun, A.F. (2023). Single-cell spatial metabolomics Jing, Y., Zuo, Y., Yu, Y., Sun, L., Yu, Z., Ma, S., Zhao, Q., Sun, G., Hu, H., Li, J., et al.
with cell-type specific protein profiling for tissue systems biology. Nat Commun 14, (2023). Single-nucleus profiling unveils a geroprotective role of the FOXO3 in
8260. primate skeletal muscle aging. Protein Cell 14, 497–512.
Hu, W., Zeng, H., Shi, Y., Zhou, C., Huang, J., Jia, L., Xu, S., Feng, X., Zeng, Y., Xiong, Joanito, I., Wirapati, P., Zhao, N., Nawaz, Z., Yeo, G., Lee, F., Eng, C.L.P., Macalinao,
T., et al. (2022). Single-cell transcriptome and translatome dual-omics reveals D.C., Kahraman, M., Srinivasan, H., et al. (2022). Single-cell and bulk
potential mechanisms of human oocyte maturation. Nat Commun 13, 5114. transcriptome sequencing identifies two epithelial tumor cell states and refines
Hu, Y., Huang, K., An, Q., Du, G., Hu, G., Xue, J., Zhu, X., Wang, C.Y., Xue, Z., and the consensus molecular classification of colorectal cancer. Nat Genet 54, 963–
Fan, G. (2016b). Simultaneous profiling of transcriptome and DNA methylome 975.
from a single cell. Genome Biol 17, 88. Johansen, N., Somasundaram, S., Travaglini, K.J., Yanny, A.M., Shumyatcher, M.,
Huang, D., Zuo, Y., Zhang, C., Sun, G., Jing, Y., Lei, J., Ma, S., Sun, S., Lu, H., Zhang, Casper, T., Cobbs, C., Dee, N., Ellenbogen, R., Ferreira, M., et al. (2023).
X., et al. (2023a). A single-nucleus transcriptomic atlas of primate testicular aging Interindividual variation in human cortical cell type abundance and expression.
reveals exhaustion of the spermatogonial stem cell reservoir and loss of Sertoli cell Science 382, eadf2359.
homeostasis. Protein Cell 14, 888–907. Johnson, W.E., Li, C., and Rabinovic, A. (2007). Adjusting batch effects in microarray
Huang, M., Wang, J., Torre, E., Dueck, H., Shaffer, S., Bonasio, R., Murray, J.I., Raj, A., expression data using empirical Bayes methods. Biostatistics 8, 118–127.
Li, M., and Zhang, N.R. (2018). SAVER: Gene expression recovery for single-cell Jordan, C.T., Guzman, M.L., and Noble, M. (2006). Cancer stem cells. N Engl J Med
RNA sequencing. Nat Methods 15, 539–542. 355, 1253–1261.
Huang, S., Wang, X., Wang, Y., Wang, Y., Fang, C., Wang, Y., Chen, S., Chen, R., Lei, Jorstad, N.L., Close, J., Johansen, N., Yanny, A.M., Barkan, E.R., Travaglini, K.J.,
T., Zhang, Y., et al. (2023b). Deciphering and advancing CAR T-cell therapy with Bertagnolli, D., Campos, J., Casper, T., Crichton, K., et al. (2023a). Transcriptomic
single-cell sequencing technologies. Mol Cancer 22, 80. cytoarchitecture reveals principles of human neocortex organization. Science 382,
Hubert, L., and Arabie, P. (1985). Comparing partitions. J Classification 2, 193–218. eadf6812.
Hwang, B., Lee, J.H., and Bang, D. (2018). Single-cell RNA sequencing technologies Jorstad, N.L., Song, J.H.T., Exposito-Alonso, D., Suresh, H., Castro-Pacheco, N.,
and bioinformatics pipelines. Exp Mol Med 50, 1–14. Krienen, F.M., Yanny, A.M., Close, J., Gelfand, E., Long, B., et al. (2023b).
Im, Y., and Kim, Y. (2023). A comprehensive overview of RNA deconvolution Comparative transcriptomics reveals human-specific cortical features. Science 382,
methods and their application. Mol Cells 46, 99–105. eade9516.
Imdahl, F., Vafadarnejad, E., Homberger, C., Saliba, A.E., and Vogel, J. (2020). Single- Jovic, D., Liang, X., Zeng, H., Lin, L., Xu, F., and Luo, Y. (2022). Single-cell RNA
cell RNA-sequencing reports growth-condition-specific global transcriptomes of sequencing technologies and applications: a brief overview. Clin Transl Med 12,
individual bacteria. Nat Microbiol 5, 1202–1206. e694.

1272 SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 [Link]
Junker, J.P., Noël, E.S., Guryev, V., Peterson, K.A., Shah, G., Huisken, J., McMahon, A. 610–620.
P., Berezikov, E., Bakkers, J., and van Oudenaarden, A. (2014). Genome-wide RNA Komatsu, J., Cico, A., Poncin, R., Le Bohec, M., Morf, J., Lipin, S., Graindorge, A.,
tomography in the zebrafish embryo. Cell 159, 662–675. Eckert, H., Saffarian, A., Cathaly, L., et al. (2023). RevGel-seq: Instrument-free
Kalucka, J., de Rooij, L.P.M.H., Goveia, J., Rohlenova, K., Dumas, S.J., Meta, E., single-cell RNA sequencing using a reversible hydrogel for cell-specific barcoding.
Conchinha, N.V., Taverna, F., Teuwen, L.A., Veys, K., et al. (2020). Single-cell Sci Rep 13, 4866.
transcriptome atlas of murine endothelial cells. Cell 180, 764–779.e20. Korsunsky, I., Millard, N., Fan, J., Slowikowski, K., Zhang, F., Wei, K., Baglaenko, Y.,
Kang, B., Camps, J., Fan, B., Jiang, H., Ibrahim, M.M., Hu, X., Qin, S., Kirchhoff, D., Brenner, M., Loh, P., and Raychaudhuri, S. (2019). Fast, sensitive and accurate
Chiang, D.Y., Wang, S., et al. (2022). Parallel single-cell and bulk transcriptome integration of single-cell data with Harmony. Nat Methods 16, 1289–1296.
analyses reveal key features of the gastric tumor microenvironment. Genome Biol Kouno, T., Moody, J., Kwon, A.T.J., Shibayama, Y., Kato, S., Huang, Y., Böttcher, M.,
23, 265. Motakis, E., Mendez, M., Severin, J., et al. (2019). C1 CAGE detects transcription
Kang, K., Meng, Q., Shats, I., Umbach, D.M., Li, M., Li, Y., Li, X., Li, L., and Przytycka, start sites and enhancer activity at single-cell resolution. Nat Commun 10, 360.
T.M. (2019). CDSeq: A novel complete deconvolution method for dissecting Kozlov, A., Alves, J.M., Stamatakis, A., and Posada, D. (2022). CellPhy: Accurate and
heterogeneous samples using gene expression data. PLoS Comput Biol 15, fast probabilistic inference of single-cell phylogenies from scDNA-seq data. Genome
e1007510. Biol 23, 37.
Kang, Y., and Pantel, K. (2013). Tumor cell dissemination: emerging biological Krishnamurty, A.T., Shyer, J.A., Thai, M., Gandham, V., Buechler, M.B., Yang, Y.A.,
insights from animal models and cancer patients. Cancer Cell 23, 573–581. Pradhan, R.N., Wang, A.W., Sanchez, P.L., Qu, Y., et al. (2022). LRRC15+
Kanton, S., Boyle, M.J., He, Z., Santel, M., Weigert, A., Sanchís-Calleja, F., Guijarro, P., myofibroblasts dictate the stromal setpoint to suppress tumour immunity. Nature
Sidow, L., Fleck, J.S., Han, D., et al. (2019). Organoid single-cell genomic atlas 611, 148–154.
uncovers human-specific features of brain development. Nature 574, 418–422. Ku, W.L., Nakamura, K., Gao, W., Cui, K., Hu, G., Tang, Q., Ni, B., and Zhao, K.
Karimi, E., Yu, M.W., Maritan, S.M., Perus, L.J.M., Rezanejad, M., Sorin, M., Dankner, (2019). Single-cell chromatin immunocleavage sequencing (scChIC-seq) to profile
M., Fallah, P., Doré, S., Zuo, D., et al. (2023). Single-cell spatial immune landscapes histone modification. Nat Methods 16, 323–325.
of primary and metastatic brain tumours. Nature 614, 555–563. Ku, W.L., Pan, L., Cao, Y., Gao, W., and Zhao, K. (2021). Profiling single-cell histone
Karras, P., Bordeu, I., Pozniak, J., Nowosad, A., Pazzi, C., Van Raemdonck, N., modifications using indexing chromatin immunocleavage sequencing. Genome Res
Landeloos, E., Van Herck, Y., Pedri, D., Bervoets, G., et al. (2022). A cellular 31, 1831–1842.
hierarchy in melanoma uncouples growth and metastasis. Nature 610, 190–198. Kuchina, A., Brettner, L.M., Paleologu, L., Roco, C.M., Rosenberg, A.B., Carignano, A.,
Kaur, R.P., Ludhiadch, A., and Munshi, A. (2019). Single-cell genomics: technology Kibler, R., Hirano, M., DePaolo, R.W., and Seelig, G. (2021). Microbial single-cell
and applications. In: Single-Cell Omics. New York: Academic Press. 179–197. RNA sequencing by split-pool barcoding. Science 371, eaba5257.
Kaya-Okur, H.S., Wu, S.J., Codomo, C.A., Pledger, E.S., Bryson, T.D., Henikoff, J.G., Lähnemann, D., Köster, J., Szczurek, E., McCarthy, D.J., Hicks, S.C., Robinson, M.D.,
Ahmad, K., and Henikoff, S. (2019). CUT&Tag for efficient epigenomic profiling of Vallejos, C.A., Campbell, K.R., Beerenwinkel, N., Mahfouz, A., et al. (2020). Eleven
small samples and single cells. Nat Commun 10, 1930. grand challenges in single-cell data science. Genome Biol 21, 31.
Ke, R., Mignardi, M., Pacureanu, A., Svedlund, J., Botling, J., Wählby, C., and Nilsson, Lai, B., Gao, W., Cui, K., Xie, W., Tang, Q., Jin, W., Hu, G., Ni, B., and Zhao, K. (2018).
M. (2013). In situ sequencing for RNA analysis in preserved tissue and cells. Nat Principles of nucleosome organization revealed by single-cell micrococcal nuclease
Methods 10, 857–860. sequencing. Nature 562, 281–285.
Kelly, T.K., Liu, Y., Lay, F.D., Liang, G., Berman, B.P., and Jones, P.A. (2012). Lake, B.B., Chen, S., Hoshi, M., Plongthongkum, N., Salamon, D., Knoten, A., Vijayan,
Genome-wide mapping of nucleosome positioning and DNA methylation within A., Venkatesh, R., Kim, E.H., Gao, D., et al. (2019). A single-nucleus RNA-
individual DNA molecules. Genome Res 22, 2497–2506. sequencing pipeline to decipher the molecular anatomy and pathophysiology of
Kelsey, G., Stegle, O., and Reik, W. (2017). Single-cell epigenomics: Recording the past human kidneys. Nat Commun 10, 2832.
and predicting the future. Science 358, 69–75. Lake, B.B., Chen, S., Sos, B.C., Fan, J., Kaeser, G.E., Yung, Y.C., Duong, T.E., Gao, D.,
Kieffer, Y., Hocine, H.R., Gentric, G., Pelon, F., Bernard, C., Bourachot, B., Lameiras, Chun, J., Kharchenko, P.V., et al. (2018). Integrative single-cell analysis of
S., Albergante, L., Bonneau, C., Guyard, A., et al. (2020). Single-cell analysis transcriptional and epigenetic states in the human adult brain. Nat Biotechnol 36,
reveals fibroblast clusters linked to immunotherapy resistance in cancer. Cancer 70–80.
Discov 10, 1330–1351. Lake, B.B., Menon, R., Winfree, S., Hu, Q., Melo Ferreira, R., Kalhor, K., Barwinska, D.,
Kim, C., Gao, R., Sei, E., Brandt, R., Hartman, J., Hatschek, T., Crosetto, N., Foukakis, Otto, E.A., Ferkowicz, M., Diep, D., et al. (2023). An atlas of healthy and injured cell
T., and Navin, N.E. (2018). Chemoresistance evolution in triple-negative breast states and niches in the human kidney. Nature 619, 585–594.
cancer delineated by single-cell sequencing. Cell 173, 879–893.e13. Lareau, C.A., Duarte, F.M., Chew, J.G., Kartha, V.K., Burkett, Z.D., Kohlway, A.S.,
Kim, C.N., Shin, D., Wang, A., and Nowakowski, T.J. (2023a). Spatiotemporal Pokholok, D., Aryee, M.J., Steemers, F.J., Lebofsky, R., et al. (2019). Droplet-based
molecular dynamics of the developing human thalamus. Science 382, eadf9941. combinatorial indexing for massive-scale single-cell chromatin accessibility. Nat
Kim, D., Kobayashi, T., Voisin, B., Jo, J.H., Sakamoto, K., Jin, S.P., Kelly, M., Pasieka, Biotechnol 37, 916–924.
H.B., Naff, J.L., Meyerle, J.H., et al. (2020). Targeted therapy guided by single-cell Lareau, C.A., Liu, V., Muus, C., Praktiknjo, S.D., Nitsch, L., Kautz, P., Sandor, K., Yin,
transcriptomic analysis in drug-induced hypersensitivity syndrome: A case report. Y., Gutierrez, J.C., Pelka, K., et al. (2023). Mitochondrial single-cell ATAC-seq for
Nat Med 26, 236–243. high-throughput multi-omic detection of mitochondrial genotypes and chromatin
Kim, S.H., Marinov, G.K., Bagdatli, S.T., Higashino, S.I., Shipony, Z., Kundaje, A., and accessibility. Nat Protoc 18, 1416–1440.
Greenleaf, W.J. (2023b). Simultaneous single-cell profiling of the transcriptome and Lareau, C.A., Ludwig, L.S., Muus, C., Gohil, S.H., Zhao, T., Chiang, Z., Pelka, K.,
accessible chromatin using SHARE-seq. Methods Mol Biol 2611, 187–230. Verboon, J.M., Luo, W., Christian, E., et al. (2021a). Massively parallel single-cell
King, H.W., Wells, K.L., Shipony, Z., Kathiria, A.S., Wagar, L.E., Lareau, C., Orban, N., mitochondrial DNA genotyping and chromatin profiling. Nat Biotechnol 39, 451–
Capasso, R., Davis, M.M., Steinmetz, L.M., et al. (2021). Integrated single-cell 461.
transcriptomics and epigenomics reveals strong germinal center-associated etiology Lareau, C.A., Parker, K.R., and Satpathy, A.T. (2021b). Charting the tumor antigen
of autoimmune risk loci. Sci Immunol 6, eabh3768. maps drawn by single-cell genomics. Cancer Cell 39, 1553–1557.
Kishi, J.Y., Lapan, S.W., Beliveau, B.J., West, E.R., Zhu, A., Sasaki, H.M., Saka, S.K., Larsson, L., Frisén, J., and Lundeberg, J. (2021). Spatially resolved transcriptomics
Wang, Y., Cepko, C.L., and Yin, P. (2019). SABER amplifies FISH: enhanced adds a new dimension to genomics. Nat Methods 18, 15–18.
multiplexed imaging of RNA and DNA in cells and tissues. Nat Methods 16, 533– Lawson, D.A., Bhakta, N.R., Kessenbrock, K., Prummel, K.D., Yu, Y., Takai, K., Zhou,
544. A., Eyob, H., Balakrishnan, S., Wang, C.Y., et al. (2015). Single-cell analysis reveals
Klein, A.M., Mazutis, L., Akartuna, I., Tallapragada, N., Veres, A., Li, V., Peshkin, L., a stem-cell program in human metastatic breast cancer cells. Nature 526, 131–
Weitz, D.A., and Kirschner, M.W. (2015). Droplet barcoding for single-cell 135.
transcriptomics applied to embryonic stem cells. Cell 161, 1187–1201. Leader, A.M., Grout, J.A., Maier, B.B., Nabet, B.Y., Park, M.D., Tabachnikova, A.,
Kleshchevnikov, V., Shmatko, A., Dann, E., Aivazidis, A., King, H.W., Li, T., Chang, C., Walker, L., Lansky, A., Le Berichel, J., et al. (2021). Single-cell analysis
Elmentaite, R., Lomakin, A., Kedlian, V., Gayoso, A., et al. (2022). Cell2location of human non-small cell lung cancer lesions refines tumor classification and patient
maps fine-grained cell types in spatial transcriptomics. Nat Biotechnol 40, 661– stratification. Cancer Cell 39, 1594–1609.e12.
671. Lee, B.R., Dalley, R., Miller, J.A., Chartrand, T., Close, J., Mann, R., Mukora, A., Ng, L.,
Koboldt, D.C., Steinberg, K.M., Larson, D.E., Wilson, R.K., and Mardis, E.R. (2013). Alfiler, L., Baker, K., et al. (2023). Signature morphoelectric properties of diverse
The next-generation sequencing revolution and its impact on genomics. Cell 155, GABAergic interneurons in the human neocortex. Science 382, eadf6484.
27–38. Lee, D.S., Luo, C., Zhou, J., Chandran, S., Rivkin, A., Bartlett, A., Nery, J.R.,
Kolodziejczyk, A.A., Kim, J.K., Svensson, V., Marioni, J.C., and Teichmann, S.A. Fitzpatrick, C., O’Connor, C., Dixon, J.R., et al. (2019). Simultaneous profiling of 3D
(2015). The technology and biology of single-cell RNA sequencing. Mol Cell 58, genome structure and DNA methylation in single human cells. Nat Methods 16,

[Link] SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 1273
 999–1006. Liao, S., Heng, Y., Liu, W., Xiang, J., Ma, Y., Chen, L., Feng, X., Jia, D., Liang, D.,
Lee, J.H., Daugharthy, E.R., Scheiman, J., Kalhor, R., Ferrante, T.C., Terry, R., Huang, C., et al. (2023). Integrated spatial transcriptomic and proteomic analysis
Turczyk, B.M., Yang, J.L., Lee, H.S., Aach, J., et al. (2015). Fluorescent in situ of fresh frozen tissue based on Stereo-seq. bioRxiv, 2023.04.28.538364.
sequencing (FISSEQ) of RNA for gene expression profiling in intact cells and tissues. Liao, Y., Ma, L., Guo, Q., Weigao, E., Fang, X., Yang, L., Ruan, F., Wang, J., Zhang, P.,
Nat Protoc 10, 442–458. Sun, Z., et al. (2022). Cell landscape of larval and adult Xenopus laevis at single-cell
Lee, J.H., Daugharthy, E.R., Scheiman, J., Kalhor, R., Yang, J.L., Ferrante, T.C., Terry, resolution. Nat Commun 13, 4306.
R., Jeanty, S.S.F., Li, C., Amamoto, R., et al. (2014). Highly multiplexed subcellular Lieberman-Aiden, E., van Berkum, N.L., Williams, L., Imakaev, M., Ragoczy, T.,
RNA sequencing in situ. Science 343, 1360–1363. Telling, A., Amit, I., Lajoie, B.R., Sabo, P.J., Dorschner, M.O., et al. (2009).
Lee, K.S., Pereira, F.C., Palatinszky, M., Behrendt, L., Alcolombri, U., Berry, D., Comprehensive mapping of long-range interactions reveals folding principles of the
Wagner, M., and Stocker, R. (2021). Optofluidic Raman-activated cell sorting for human genome. Science 326, 289–293.
targeted genome retrieval or cultivation of microbial cells with specific functions. Lim, J., Chin, V., Fairfax, K., Moutinho, C., Suan, D., Ji, H., and Powell, J.E. (2023).
Nat Protoc 16, 634–676. Transitioning single-cell genomics into the clinic. Nat Rev Genet 24, 573–584.
Leighton, J., Hu, M., Sei, E., Meric-Bernstam, F., and Navin, N.E. (2023). Lin, D., Shen, L., Luo, M., Zhang, K., Li, J., Yang, Q., Zhu, F., Zhou, D., Zheng, S., Chen,
Reconstructing mutational lineages in breast cancer by multi-patient-targeted Y., et al. (2021a). Circulating tumor cells: biology and clinical significance. Sig
single-cell DNA sequencing. Cell Genomics 3, 100215. Transduct Target Ther 6, 404.
Lein, E., Borm, L.E., and Linnarsson, S. (2017). The promise of spatial transcriptomics Lin, T.Y., Gerber, T., Taniguchi-Sugiura, Y., Murawala, P., Hermann, S., Grosser, L.,
for neuroscience in the era of molecular cell typing. Science 358, 64–69. Shibata, E., Treutlein, B., and Tanaka, E.M. (2021b). Fibroblast dedifferentiation as
Li, B., Li, T., Liu, J.S., and Liu, X.S. (2020). Computational deconvolution of tumor- a determinant of successful regeneration. Dev Cell 56, 1541–1551.e6.
infiltrating immune components with bulk tumor gene expression data. Methods Lin, Y., Ghazanfar, S., Wang, K.Y.X., Gagnon-Bartsch, J.A., Lo, K.K., Su, X., Han, Z.G.,
Mol Biol 2120, 249–262. Ormerod, J.T., Speed, T.P., Yang, P., et al. (2019). scMerge leverages factor
Li, G., Liu, Y., Zhang, Y., Kubo, N., Yu, M., Fang, R., Kellis, M., and Ren, B. (2019). analysis, stable expression, and pseudoreplication to merge multiple single-cell
Joint profiling of DNA methylation and chromatin architecture in single cells. Nat RNA-seq datasets. Proc Natl Acad Sci USA 116, 9775–9784.
Methods 16, 991–993. Lindeboom, R.G.H., Regev, A., and Teichmann, S.A. (2021). Towards a human cell
Li, H., Janssens, J., De Waegeneer, M., Kolluru, S.S., Davie, K., Gardeux, V., Saelens, atlas: taking notes from the past. Trends Genet 37, 625–630.
W., David, F.P.A., Brbić, M., Spanier, K., et al. (2022a). Fly Cell Atlas: a single- Linderman, G.C., Zhao, J., Roulis, M., Bielecki, P., Flavell, R.A., Nadler, B., and Kluger,
nucleus transcriptomic atlas of the adult fruit fly. Science 375, eabk2432. Y. (2022). Zero-preserving imputation of single-cell RNA-seq data. Nat Commun
Li, H., Ma, T., Hao, M., Guo, W., Gu, J., Zhang, X., and Wei, L. (2023a). Decoding 13, 192.
functional cell-cell communication events by multi-view graph learning on spatial Ling, A.L., Solomon, I.H., Landivar, A.M., Nakashima, H., Woods, J.K., Santos, A.,
transcriptomics. Brief Bioinf 24, bbad359. Masud, N., Fell, G., Mo, X., Yilmaz, A.S., et al. (2023). Clinical trial links oncolytic
Li, H., Sun, Y., Hong, H., Huang, X., Tao, H., Huang, Q., Wang, L., Xu, K., Gan, J., immunoactivation to survival in glioblastoma. Nature 623, 157–166.
Chen, H., et al. (2022b). Inferring transcription factor regulatory networks from Liscovitch-Brauer, N., Montalbano, A., Deng, J., Méndez-Mancilla, A., Wessels, H.H.,
single-cell ATAC-seq data based on graph neural networks. Nat Mach Intell 4, Moss, N.G., Kung, C.Y., Sookdeo, A., Guo, X., Geller, E., et al. (2021). Profiling the
389–400. genetic determinants of chromatin accessibility with scalable single-cell CRISPR
Li, J., Wang, J., Zhang, P., Wang, R., Mei, Y., Sun, Z., Fei, L., Jiang, M., Ma, L., E, W., et screens. Nat Biotechnol 39, 1270–1277.
al. (2022c). Deep learning of cross-species single-cell landscapes identifies Litviňuková, M., Talavera-López, C., Maatz, H., Reichart, D., Worth, C.L., Lindberg, E.
conserved regulatory programs underlying cell types. Nat Genet 54, 1711–1720. L., Kanda, M., Polanski, K., Heinig, M., Lee, M., et al. (2020). Cells of the adult
Li, J., Zhang, Z., Song, R., Li, Y., and Du, Q. (2024). SCFormer: spectral coordinate human heart. Nature 588, 466–472.
transformer for cross-domain few-shot hyperspectral image classification. IEEE Liu, B., Hu, X., Feng, K., Gao, R., Xue, Z., Zhang, S., Zhang, Y., Corse, E., Hu, Y., Han,
Trans Image Process 33, 840–855. W., et al. (2022a). Temporal single-cell tracing reveals clonal revival and
Li, J., Zheng, Y., Yan, P., Song, M., Wang, S., Sun, L., Liu, Z., Ma, S., Izpisua Belmonte, expansion of precursor exhausted T cells during anti-PD-1 therapy in lung cancer.
J.C., Chan, P., et al. (2021a). A single-cell transcriptomic atlas of primate Nat Cancer 3, 108–121.
pancreatic islet aging. Natl Sci Rev 8, nwaa127. Liu, B., Zhang, Y., Wang, D., Hu, X., and Zhang, Z. (2022b). Single-cell meta-analyses
Li, L., Dong, J., Yan, L., Yong, J., Liu, X., Hu, Y., Fan, X., Wu, X., Guo, H., Wang, X., et reveal responses of tumor-reactive CXCL13+ T cells to immune-checkpoint
al. (2017). Single-cell RNA-Seq analysis maps development of human germline blockade. Nat Cancer 3, 1123–1136.
cells and gonadal niche interactions. Cell Stem Cell 20, 858–873.e4. Liu, C., Li, X., Huang, Q., Zhang, M., Lei, T., Wang, F., Zou, W., Huang, R., Hu, X.,
Li, L., Guo, F., Gao, Y., Ren, Y., Yuan, P., Yan, L., Li, R., Lian, Y., Li, J., Hu, B., et al. Wang, C., et al. (2023a). Single-cell RNA-sequencing reveals radiochemotherapy-
(2018). Single-cell multi-omics sequencing of human early embryos. Nat Cell Biol induced innate immune activation and MHC-II upregulation in cervical cancer. Sig
20, 847–858. Transduct Target Ther 8, 44.
Li, R., Wang, Z., Wang, J.W., and Li, L. (2023b). Combining single-cell RNA Liu, F., Wang, Y., Gu, H., and Wang, X. (2023b). Technologies and applications of
sequencing with spatial transcriptome analysis reveals dynamic molecular maps of single-cell DNA methylation sequencing. Theranostics 13, 2439–2454.
cambium differentiation in the primary and secondary growth of trees. Plant Liu, H., Zhou, J., Tian, W., Luo, C., Bartlett, A., Aldridge, A., Lucero, J., Osteen, J.K.,
Commun 4, 100665. Nery, J.R., Chen, H., et al. (2021). DNA methylation atlas of the mouse brain at
Li, W., Calder, R.B., Mar, J.C., and Vijg, J. (2015). Single-cell transcriptogenomics single-cell resolution. Nature 598, 120–128.
reveals transcriptional exclusion of ENU-mutated alleles. Mutat Res 772, 55–62. Liu, J., Gao, C., Sodicoff, J., Kozareva, V., Macosko, E.Z., and Welch, J.D. (2020a).
Li, W., Lu, J., Lu, P., Gao, Y., Bai, Y., Chen, K., Su, X., Li, M., Liu, J., Chen, Y., et al. Jointly defining cell types from multiple single-cell datasets using LIGER. Nat Protoc
(2023c). scNanoHi-C: a single-cell long-read concatemer sequencing method to 15, 3632–3662.
reveal high-order chromatin structures within individual cells. Nat Methods 20, Liu, L., Liu, C., Quintero, A., Wu, L., Yuan, Y., Wang, M., Cheng, M., Leng, L., Xu, L.,
1493–1505. Dong, G., et al. (2019). Deconvolution of single-cell multi-omics layers reveals
Li, W.V., and Li, J.J. (2018). An accurate and robust imputation method scImpute for regulatory heterogeneity. Nat Commun 10, 470.
single-cell RNA-seq data. Nat Commun 9, 997. Liu, X., Mao, D., Song, Y., Zhu, L., Isak, A.N., Lu, C., Deng, G., Chen, F., Sun, F., Yang,
Li, X., and Wang, C.Y. (2021). From bulk, single-cell to spatial RNA sequencing. Int J Y., et al. (2022c). Computer-aided design of reversible hybridization chain reaction
Oral Sci 13, 36. (CAD-HCR) enables multiplexed single-cell spatial proteomics imaging. Sci Adv 8,
Li, Y., Huang, Z., Zhang, Z., Wang, Q., Li, F., Wang, S., Ji, X., Shu, S., Fang, X., and eabk0133.
Jiang, L. (2023d). FIPRESCI: Droplet microfluidics based combinatorial indexing for Liu, X., Shen, Q., and Zhang, S. (2023c). Cross-species cell-type assignment from
massive-scale 5′-end single-cell RNA sequencing. Genome Biol 24, 70. single-cell RNA-seq data by a heterogeneous graph neural network. Genome Res
Li, Y., Li, H., Xie, Y., Chen, S., Qin, R., Dong, H., Yu, Y., Wang, J., Qian, X., and Qin, 33, 96–111.
W. (2021b). An integrated strategy for mass spectrometry-based multiomics Liu, X., Song, J., Zhang, H., Liu, X., Zuo, F., Zhao, Y., Zhao, Y., Yin, X., Guo, X., Wu,
analysis of single cells. Anal Chem 93, 14059–14067. X., et al. (2023d). Immune checkpoint HLA-E: CD94-NKG2A mediates evasion of
Li, Y., Li, Z., Wang, C., Yang, M., He, Z., Wang, F., Zhang, Y., Li, R., Gong, Y., Wang, circulating tumor cells from NK cell surveillance. Cancer Cell 41, 272–287.e9.
B., et al. (2023e). Spatiotemporal transcriptome atlas reveals the regional Liu, Y., DiStasio, M., Su, G., Asashima, H., Enninful, A., Qin, X., Deng, Y., Nam, J.,
specification of the developing human brain. Cell 186, 5892–5909.e22. Gao, F., Bordignon, P., et al. (2023e). High-plex protein and whole transcriptome
Li, Y.E., Preissl, S., Miller, M., Johnson, N.D., Wang, Z., Jiao, H., Zhu, C., Wang, Z., Xie, co-mapping at cellular resolution with spatial CITE-seq. Nat Biotechnol 41, 1405–
Y., Poirion, O., et al. (2023f). A comparative atlas of single-cell chromatin 1409.
accessibility in the human brain. Science 382, eadf7044. Liu, Y., Wang, H., Taylor, M., Cook, C., Martínez-Berdeja, A., North, J.P., Harirchian,

1274 SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 [Link]
 P., Hailer, A.A., Zhao, Z., Ghadially, R., et al. (2022d). Classification of human Cell Biol 23, 303–304.
chronic inflammatory skin disease based on single-cell immune profiling. Sci Ma, S., Sun, S., Li, J., Fan, Y., Qu, J., Sun, L., Wang, S., Zhang, Y., Yang, S., Liu, Z., et
Immunol 7, eabl9165. al. (2021). Single-cell transcriptomic atlas of primate cardiopulmonary aging. Cell
Liu, Y., Yang, M., Deng, Y., Su, G., Enninful, A., Guo, C.C., Tebaldi, T., Zhang, D., Kim, Res 31, 415–432.
D., Bai, Z., et al. (2020b). High-spatial-resolution multi-omics sequencing via Ma, S., Zhang, B., LaFave, L.M., Earl, A.S., Chiang, Z., Hu, Y., Ding, J., Brack, A.,
deterministic barcoding in tissue. Cell 183, 1665–1681.e18. Kartha, V.K., Tay, T., et al. (2020b). Chromatin potential identified by shared
Liu, Z., Chen, Y., Xia, Q., Liu, M., Xu, H., Chi, Y., Deng, Y., and Xing, D. (2023f). single-cell profiling of RNA and chromatin. Cell 183, 1103–1116.e20.
Linking genome structures to functions by simultaneous single-cell Hi-C and RNA- Ma, Y., and Zhou, X. (2022). Spatially informed cell-type deconvolution for spatial
seq. Science 380, 1070–1076. transcriptomics. Nat Biotechnol 40, 1349–1359.
Llorens-Bobadilla, E., Zamboni, M., Marklund, M., Bhalla, N., Chen, X., Hartman, J., Macarron, R., Banks, M.N., Bojanic, D., Burns, D.J., Cirovic, D.A., Garyantes, T.,
Frisén, J., and Ståhl, P.L. (2023). Solid-phase capture and profiling of open Green, D.V.S., Hertzberg, R.P., Janzen, W.P., Paslay, J.W., et al. (2011). Impact of
chromatin by spatial ATAC. Nat Biotechnol 41, 1085–1088. high-throughput screening in biomedical research. Nat Rev Drug Discov 10, 188–
Lochs, S.J.A., van der Weide, R.H., de Luca, K.L., Korthout, T., van Beek, R.E., Kimura, 195.
H., and Kind, J. (2024). Combinatorial single-cell profiling of major chromatin types Macaulay, I.C., Haerty, W., Kumar, P., Li, Y.I., Hu, T.X., Teng, M.J., Goolam, M.,
with MAbID. Nat Methods 21, 72–82. Saurat, N., Coupland, P., Shirley, L.M., et al. (2015). G&T-seq: Parallel sequencing
Lopez, R., Regier, J., Cole, M.B., Jordan, M.I., and Yosef, N. (2018). Deep generative of single-cell genomes and transcriptomes. Nat Methods 12, 519–522.
modeling for single-cell transcriptomics. Nat Methods 15, 1053–1058. Macosko, E.Z., Basu, A., Satija, R., Nemesh, J., Shekhar, K., Goldman, M., Tirosh, I.,
Lotfollahi, M., Wolf, F.A., and Theis, F.J. (2019). scGen predicts single-cell Bialas, A.R., Kamitaki, N., Martersteck, E.M., et al. (2015). Highly parallel genome-
perturbation responses. Nat Methods 16, 715–721. wide expression profiling of individual cells using nanoliter droplets. Cell 161,
Lovatt, D., Ruble, B.K., Lee, J., Dueck, H., Kim, T.K., Fisher, S., Francis, C., Spaethling, 1202–1214.
J.M., Wolf, J.A., Grady, M.S., et al. (2014). Transcriptome in vivo analysis (TIVA) of Maden, S.K., Kwon, S.H., Huuki-Myers, L.A., Collado-Torres, L., Hicks, S.C., and
spatially defined single cells in live tissue. Nat Methods 11, 190–196. Maynard, K.R. (2023). Challenges and opportunities to computationally decon-
Lu, Q., Liu, J., Zhao, S., Gomez Castro, M.F., Laurent-Rolle, M., Dong, J., Ran, X., volve heterogeneous tissue with varying cell sizes using single-cell RNA-sequencing
Damani-Yokota, P., Tang, H., Karakousi, T., et al. (2021). SARS-CoV-2 exacerbates datasets. Genome Biol 24, 288.
proinflammatory responses in myeloid cells through C-type lectin receptors and Magen, A., Hamon, P., Fiaschi, N., Soong, B.Y., Park, M.D., Mattiuz, R., Humblin, E.,
Tweety family member 2. Immunity 54, 1304–1319.e9. Troncoso, L., D’souza, D., Dawson, T., et al. (2023). Intratumoral dendritic cell-
Lu, T., Ang, C.E., and Zhuang, X. (2022). Spatially resolved epigenomic profiling of CD4+ T helper cell niches enable CD8+ T cell differentiation following PD-1
single cells in complex tissues. Cell 185, 4448–4464.e17. blockade in hepatocellular carcinoma. Nat Med 29, 1389–1399.
Lu, T.C., Brbić, M., Park, Y.J., Jackson, T., Chen, J., Kolluru, S.S., Qi, Y., Katheder, N. Mah, J.L., and Dunn, C.W. (2024). Cell type evolution reconstruction across species
S., Cai, X.T., Lee, S., et al. (2023). Aging Fly Cell Atlas identifies exhaustive aging through cell phylogenies of single-cell RNA sequencing data. Nat Ecol Evol 8, 325–
features at cellular resolution. Science 380, eadg0934. 338.
Lu, Y., Zhao, Q., Liao, J.Y., Song, E., Xia, Q., Pan, J., Li, Y., Li, J., Zhou, B., Ye, Y., et al. Mahat, D.B., Tippens, N.D., Martin-Rufino, J.D., Waterton, S.K., Fu, J., Blatt, S.E., and
(2020). Complement signals determine opposite effects of B cells in chemotherapy- Sharp, P.A. (2024). Single-cell nascent RNA sequencing unveils coordinated global
induced immunity. Cell 180, 1081–1097.e24. transcription. Nature 631, 216–223.
Lubeck, E., and Cai, L. (2012). Single-cell systems biology by super-resolution imaging Maier, B., Leader, A.M., Chen, S.T., Tung, N., Chang, C., LeBerichel, J., Chudnovskiy,
and combinatorial labeling. Nat Methods 9, 743–748. A., Maskey, S., Walker, L., Finnigan, J.P., et al. (2020). A conserved dendritic-cell
Luecken, M.D., Büttner, M., Chaichoompu, K., Danese, A., Interlandi, M., Mueller, M. regulatory program limits antitumour immunity. Nature 580, 257–262.
F., Strobl, D.C., Zappia, L., Dugas, M., Colomé-Tatché, M., et al. (2022). Markham, A.F. (1993). The polymerase chain reaction: a tool for molecular medicine.
Benchmarking atlas-level data integration in single-cell genomics. Nat Methods BMJ 306, 441–446.
19, 41–50. Markodimitraki, C.M., Rang, F.J., Rooijers, K., de Vries, S.S., Chialastri, A., de Luca, K.
Luo, C., Keown, C.L., Kurihara, L., Zhou, J., He, Y., Li, J., Castanon, R., Lucero, J., L., Lochs, S.J.A., Mooijman, D., Dey, S.S., and Kind, J. (2020). Simultaneous
Nery, J.R., Sandoval, J.P., et al. (2017). Single-cell methylomes identify neuronal quantification of protein–DNA interactions and transcriptomes in single cells with
subtypes and regulatory elements in mammalian cortex. Science 357, 600–604. scDam&T-seq. Nat Protoc 15, 1922–1953.
Luo, C., Liu, H., Xie, F., Armand, E.J., Siletti, K., Bakken, T.E., Fang, R., Doyle, W.I., Martin, J.C., Chang, C., Boschetti, G., Ungaro, R., Giri, M., Grout, J.A., Gettler, K.,
Stuart, T., Hodge, R.D., et al. (2022). Single nucleus multi-omics identifies human Chuang, L., Nayar, S., Greenstein, A.J., et al. (2019). Single-cell analysis of Crohn’s
cortical cell regulatory genome diversity. Cell Genomics 2, 100107. disease lesions identifies a pathogenic cellular module associated with resistance to
Luo, C., Rivkin, A., Zhou, J., Sandoval, J.P., Kurihara, L., Lucero, J., Castanon, R., anti-TNF therapy. Cell 178, 1493–1508.e20.
Nery, J.R., Pinto-Duarte, A., Bui, B., et al. (2018). Robust single-cell DNA Martinelli, A.L., and Rapsomaniki, M.A. (2022). ATHENA: Analysis of tumor
methylome profiling with snmC-seq2. Nat Commun 9, 3824. heterogeneity from spatial omics measurements. Bioinformatics 38, 3151–3153.
Luo, Z., Wen, X., Zhao, W., Nguyen, T.C., Wan, X., Richard, J.L.C., and Zhong, S. Martins, F., Sofiya, L., Sykiotis, G.P., Lamine, F., Maillard, M., Fraga, M., Shabafrouz,
(2023). Joint profiling of multiplex chromatin interactions, gene expression, and K., Ribi, C., Cairoli, A., Guex-Crosier, Y., et al. (2019). Adverse effects of immune-
RNA-chromatin associations in single cells of the human brain. bioRxiv, checkpoint inhibitors: epidemiology, management and surveillance. Nat Rev Clin
2023.06.28.546457. Oncol 16, 563–580.
Luoma, A.M., Suo, S., Williams, H.L., Sharova, T., Sullivan, K., Manos, M., Bowling, Marx, V. (2021). Method of the year: spatially resolved transcriptomics. Nat Methods
P., Hodi, F.S., Rahma, O., Sullivan, R.J., et al. (2020). Molecular pathways of colon 18, 9–14.
inflammation induced by cancer immunotherapy. Cell 182, 655–671.e22. McGinnis, C.S., Patterson, D.M., Winkler, J., Conrad, D.N., Hein, M.Y., Srivastava, V.,
Luquette, L.J., Bohrson, C.L., Sherman, M.A., and Park, P.J. (2019). Identification of Hu, J.L., Murrow, L.M., Weissman, J.S., Werb, Z., et al. (2019). MULTI-seq: sample
somatic mutations in single cell DNA-seq using a spatial model of allelic imbalance. multiplexing for single-cell RNA sequencing using lipid-tagged indices. Nat
Nat Commun 10, 3908. Methods 16, 619–626.
Lv, J., Wei, Y., Yin, J.H., Chen, Y.P., Zhou, G.Q., Wei, C., Liang, X.Y., Zhang, Y., McKellar, D.W., Mantri, M., Hinchman, M.M., Parker, J.S.L., Sethupathy, P.,
Zhang, C.J., He, S.W., et al. (2023). The tumor immune microenvironment of Cosgrove, B.D., and De Vlaminck, I. (2023). Spatial mapping of the total
nasopharyngeal carcinoma after gemcitabine plus cisplatin treatment. Nat Med 29, transcriptome by in situ polyadenylation. Nat Biotechnol 41, 513–520.
1424–1436. McNulty, R., Sritharan, D., Pahng, S.H., Meisch, J.P., Liu, S., Brennan, M.A., Saxer, G.,
Ma, A., Wang, C., Chang, Y., Brennan, F.H., McDermaid, A., Liu, B., Zhang, C., Hormoz, S., and Rosenthal, A.Z. (2023). Probe-based bacterial single-cell RNA
Popovich, P.G., and Ma, Q. (2020a). IRIS3: Integrated cell-type-specific regulon sequencing predicts toxin regulation. Nat Microbiol 8, 934–945.
inference server from single-cell RNA-Seq. Nucleic Acids Res 48, W275–W286. Meers, M.P., Llagas, G., Janssens, D.H., Codomo, C.A., and Henikoff, S. (2023).
Ma, C., Yang, C., Peng, A., Sun, T., Ji, X., Mi, J., Wei, L., Shen, S., and Feng, Q. Multifactorial profiling of epigenetic landscapes at single-cell resolution using
(2023a). Pan-cancer spatially resolved single-cell analysis reveals the crosstalk MulTI-Tag. Nat Biotechnol 41, 708–716.
between cancer-associated fibroblasts and tumor microenvironment. Mol Cancer Menden, K., Marouf, M., Oller, S., Dalmia, A., Magruder, D.S., Kloiber, K., Heutink, P.,
22, 170. and Bonn, S. (2020). Deep learning-based cell composition analysis from tissue
Ma, P., Amemiya, H.M., He, L.L., Gandhi, S.J., Nicol, R., Bhattacharyya, R.P., Smillie, expression profiles. Sci Adv 6, eaba2619.
C.S., and Hung, D.T. (2023b). Bacterial droplet-based single-cell RNA-seq reveals Meric-Bernstam, F., Larkin, J., Tabernero, J., and Bonini, C. (2021). Enhancing anti-
antibiotic-associated heterogeneous cellular states. Cell 186, 877–891.e14. tumour efficacy with immunotherapy combinations. Lancet 397, 1010–1022.
Ma, Q., and Xu, D. (2022). Deep learning shapes single-cell data analysis. Nat Rev Mol Merritt, C.R., Ong, G.T., Church, S.E., Barker, K., Danaher, P., Geiss, G., Hoang, M.,

[Link] SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 1275
 Jung, J., Liang, Y., McKay-Fleisch, J., et al. (2020). Multiplex digital spatial profiling checkpoint blockade: response evaluation and biomarker development. Nat Rev
of proteins and RNA in fixed tissue. Nat Biotechnol 38, 586–599. Clin Oncol 14, 655–668.
Metzker, M.L. (2010). Sequencing technologies—the next generation. Nat Rev Genet Niu, M., Cao, W., Wang, Y., Zhu, Q., Luo, J., Wang, B., Zheng, H., Weitz, D.A., and
11, 31–46. Zong, C. (2023). Droplet-based transcriptome profiling of individual synapses. Nat
Mezger, A., Klemm, S., Mann, I., Brower, K., Mir, A., Bostick, M., Farmer, A., Fordyce, Biotechnol 41, 1332–1344.
P., Linnarsson, S., and Greenleaf, W. (2018). High-throughput chromatin Nolan, E., Lindeman, G.J., and Visvader, J.E. (2023). Deciphering breast cancer: from
accessibility profiling at single-cell resolution. Nat Commun 9, 3647. biology to the clinic. Cell 186, 1708–1728.
Mimitou, E.P., Cheng, A., Montalbano, A., Hao, S., Stoeckius, M., Legut, M., Roush, T., O’Connor, O.M., Alnahhas, R.N., Lugagne, J.B., and Dunlop, M.J. (2022). DeLTA 2.0:
Herrera, A., Papalexi, E., Ouyang, Z., et al. (2019). Multiplexed detection of a deep learning pipeline for quantifying single-cell spatial and temporal dynamics.
proteins, transcriptomes, clonotypes and CRISPR perturbations in single cells. Nat PLoS Comput Biol 18, e1009797.
Methods 16, 409–412. Oliveira, G., Stromhaug, K., Klaeger, S., Kula, T., Frederick, D.T., Le, P.M., Forman, J.,
Minoura, K., Abe, K., Nam, H., Nishikawa, H., and Shimamura, T. (2021). A mixture- Huang, T., Li, S., Zhang, W., et al. (2021). Phenotype, specificity and avidity of
of-experts deep generative model for integrated analysis of single-cell multiomics antitumour CD8+ T cells in melanoma. Nature 596, 119–125.
data. Cell Rep Methods 1, 100071. Packer, J.S., Zhu, Q., Huynh, C., Sivaramakrishnan, P., Preston, E., Dueck, H.,
Mitra, R.D., Butty, V.L., Shendure, J., Williams, B.R., Housman, D.E., and Church, G. Stefanik, D., Tan, K., Trapnell, C., Kim, J., et al. (2019). A lineage-resolved
M. (2003). Digital genotyping and haplotyping with polymerase colonies. Proc Natl molecular atlas of C. elegans embryogenesis at single-cell resolution. Science 365,
Acad Sci USA 100, 5926–5931. 1265.
Mitra, R.D., and Church, G.M. (1999). In situ localized amplification and contact Palla, G., Fischer, D.S., Regev, A., and Theis, F.J. (2022a). Spatial components of
replication of many individual DNA molecules. Nucleic Acids Res 27, 34e–34. molecular tissue biology. Nat Biotechnol 40, 308–318.
Mittnenzweig, M., Mayshar, Y., Cheng, S., Ben-Yair, R., Hadas, R., Rais, Y., Chomsky, Palla, G., Spitzer, H., Klein, M., Fischer, D., Schaar, A.C., Kuemmerle, L.B., Rybakov,
E., Reines, N., Uzonyi, A., Lumerman, L., et al. (2021). A single-embryo, single-cell S., Ibarra, I.L., Holmberg, O., Virshup, I., et al. (2022b). Squidpy: a scalable
time-resolved model for mouse gastrulation. Cell 184, 2825–2842.e22. framework for spatial omics analysis. Nat Methods 19, 171–178.
Moffitt, J.R., Hao, J., Wang, G., Chen, K.H., Babcock, H.P., and Zhuang, X. (2016). Pan, J., Li, Y., Lin, Z., Lan, Q., Chen, H., Zhai, M., Sui, S., Zhang, G., Cheng, Y., Tang,
High-throughput single-cell gene-expression profiling with multiplexed error- Y., et al. (2023). Spatiotemporal transcriptome atlas of human embryos after
robust fluorescence in situ hybridization. Proc Natl Acad Sci USA 113, 11046– gastrulation. bioRxiv, 2023.2004.2022.537909.
11051. Pan, L., Ku, W.L., Tang, Q., Cao, Y., and Zhao, K. (2022). scPCOR-seq enables co-
Moffitt, J.R., and Zhuang, X. (2016). RNA imaging with multiplexed error-robust profiling of chromatin occupancy and RNAs in single cells. Commun Biol 5, 678.
fluorescence in situ hybridization (MERFISH). Methods Enzymol 572, 1–49. Pang, M., and Tegnér, J. (2020). Multitask learning for transformers with application
Monaco, G., Lee, B., Xu, W., Mustafah, S., Hwang, Y.Y., Carré, C., Burdin, N., Visan, to large-scale single-cell transcriptomes. bioRxiv, 2020.2002.2005.935239.
L., Ceccarelli, M., Poidinger, M., et al. (2019). RNA-Seq signatures normalized by Papalexi, E., Mimitou, E.P., Butler, A.W., Foster, S., Bracken, B., Mauck Iii, W.M.,
mRNA abundance allow absolute deconvolution of human immune cell types. Cell Wessels, H.H., Hao, Y., Yeung, B.Z., Smibert, P., et al. (2021). Characterizing the
Rep 26, 1627–1640.e7. molecular regulation of inhibitory immune checkpoints with multimodal single-cell
Mongia, A., Sengupta, D., and Majumdar, A. (2019). McImpute: matrix completion screens. Nat Genet 53, 322–331.
based imputation for single cell RNA-seq data. Front Genet 10, 9. Patruno, L., Maspero, D., Craighero, F., Angaroni, F., Antoniotti, M., and Graudenzi,
Morris, J.A., Caragine, C., Daniloski, Z., Domingo, J., Barry, T., Lu, L., Davis, K., Ziosi, A. (2021). A review of computational strategies for denoising and imputation of
M., Glinos, D.A., Hao, S., et al. (2023). Discovery of target genes and pathways at single-cell transcriptomic data. Brief Bioinf 22, bbaa222.
GWAS loci by pooled single-cell CRISPR screens. Science 380, eadh7699. Pellegrino, M., Sciambi, A., Treusch, S., Durruthy-Durruthy, R., Gokhale, K., Jacob, J.,
Moses, L., and Pachter, L. (2022). Museum of spatial transcriptomics. Nat Methods Chen, T.X., Geis, J.A., Oldham, W., Matthews, J., et al. (2018). High-throughput
19, 534–546. single-cell DNA sequencing of acute myeloid leukemia tumors with droplet
Mukherjee, P.K., Nguyen, Q.T., Li, J., Zhao, S., Christensen, S.M., West, G.A., microfluidics. Genome Res 28, 1345–1352.
Chandra, J., Gordon, I.O., Lin, S., Wang, J., et al. (2023). Stricturing Crohn’s Peng, G., Suo, S., Chen, J., Chen, W., Liu, C., Yu, F., Wang, R., Chen, S., Sun, N., Cui,
disease single-cell RNA sequencing reveals fibroblast heterogeneity and inter- G., et al. (2016). Spatial transcriptome for the molecular annotation of lineage fates
cellular interactions. Gastroenterology 165, 1180–1196. and cell identity in mid-gastrula mouse embryo. Dev Cell 36, 681–697.
Mulder, K., Patel, A.A., Kong, W.T., Piot, C., Halitzki, E., Dunsmore, G., Khalilnezhad, Peng, G., Suo, S., Cui, G., Yu, F., Wang, R., Chen, J., Chen, S., Liu, Z., Chen, G., Qian,
S., Irac, S.E., Dubuisson, A., Chevrier, M., et al. (2021). Cross-tissue single-cell Y., et al. (2019a). Molecular architecture of lineage allocation and tissue
landscape of human monocytes and macrophages in health and disease. Immunity organization in early mouse embryo. Nature 572, 528–532.
54, 1883–1900.e5. Peng, S., Petersen, J.L., Bellone, R.R., Kalbfleisch, T., Kingsley, N.B., Barber, A.M.,
Mulqueen, R.M., Pokholok, D., O’Connell, B.L., Thornton, C.A., Zhang, F., O’Roak, B. Cappelletti, E., Giulotto, E., and Finno, C.J. (2021). Decoding the equine genome:
J., Link, J., Yardımcı, G.G., Sears, R.C., Steemers, F.J., et al. (2021). High-content lessons from ENCODE. Genes 12, 1707.
single-cell combinatorial indexing. Nat Biotechnol 39, 1574–1580. Peng, X.L., Moffitt, R.A., Torphy, R.J., Volmar, K.E., and Yeh, J.J. (2019b). De novo
Nadappuram, B.P., Cadinu, P., Barik, A., Ainscough, A.J., Devine, M.J., Kang, M., compartment deconvolution and weight estimation of tumor samples using
Gonzalez-Garcia, J., Kittler, J.T., Willison, K.R., Vilar, R., et al. (2019). Nanoscale DECODER. Nat Commun 10, 4729.
tweezers for single-cell biopsies. Nat Nanotech 14, 80–88. Penkov, D., Zubkova, E., and Parfyonova, Y. (2023). Tn5 DNA transposase in multi-
Nagano, T., Lubling, Y., Stevens, T.J., Schoenfelder, S., Yaffe, E., Dean, W., Laue, E.D., omics research. Methods Protoc 6, 24.
Tanay, A., and Fraser, P. (2013). Single-cell Hi-C reveals cell-to-cell variability in Pensold, D., and Zimmer-Bensch, G. (2020). Methods for single-cell isolation and
chromosome structure. Nature 502, 59–64. preparation. In: Yu, B., Zhang, J., Zeng, Y., Li, L., and Wang, X., eds. Single-cell
Nathan, A., Asgari, S., Ishigaki, K., Valencia, C., Amariuta, T., Luo, Y., Beynor, J.I., Sequencing and Methylation. Advances in Experimental Medicine and Biology.
Baglaenko, Y., Suliman, S., Price, A.L., et al. (2022). Single-cell eQTL models reveal Singapore: Springer. 7–27.
dynamic T cell state dependence of disease loci. Nature 606, 120–128. Perez, R.K., Gordon, M.G., Subramaniam, M., Kim, M.C., Hartoularos, G.C., Targ, S.,
Newman, A.M., Liu, C.L., Green, M.R., Gentles, A.J., Feng, W., Xu, Y., Hoang, C.D., Sun, Y., Ogorodnikov, A., Bueno, R., Lu, A., et al. (2022). Single-cell RNA-seq
Diehn, M., and Alizadeh, A.A. (2015). Robust enumeration of cell subsets from reveals cell type-specific molecular and genetic associations to lupus. Science 376,
tissue expression profiles. Nat Methods 12, 453–457. eabf1970.
Newman, A.M., Steen, C.B., Liu, C.L., Gentles, A.J., Chaudhuri, A.A., Scherer, F., Picelli, S., Björklund, Å.K., Faridani, O.R., Sagasser, S., Winberg, G., and Sandberg, R.
Khodadoust, M.S., Esfahani, M.S., Luca, B.A., Steiner, D., et al. (2019). Determining (2013). Smart-seq2 for sensitive full-length transcriptome profiling in single cells.
cell type abundance and expression from bulk tissues with digital cytometry. Nat Nat Methods 10, 1096–1098.
Biotechnol 37, 773–782. Pierce, S.E., Granja, J.M., and Greenleaf, W.J. (2021). High-throughput single-cell
Nguyen, L.T., Schmidt, H.A., von Haeseler, A., and Minh, B.Q. (2015). IQ-TREE: a fast chromatin accessibility CRISPR screens enable unbiased identification of regulatory
and effective stochastic algorithm for estimating maximum-likelihood phylogenies. networks in cancer. Nat Commun 12, 2969.
Mol Biol Evol 32, 268–274. Pijuan-Sala, B., Griffiths, J.A., Guibentif, C., Hiscock, T.W., Jawaid, W., Calero-Nieto,
Nichterwitz, S., Chen, G., Aguila Benitez, J., Yilmaz, M., Storvall, H., Cao, M., F.J., Mulas, C., Ibarra-Soria, X., Tyser, R.C.V., Ho, D.L.L., et al. (2019). A single-cell
Sandberg, R., Deng, Q., and Hedlund, E. (2016). Laser capture microscopy coupled molecular map of mouse gastrulation and early organogenesis. Nature 566, 490–
with Smart-seq2 for precise spatial transcriptomic profiling. Nat Commun 7, 495.
12139. Pisetsky, D.S. (2023). Pathogenesis of autoimmune disease. Nat Rev Nephrol 19,
Nishino, M., Ramaiya, N.H., Hatabu, H., and Hodi, F.S. (2017). Monitoring immune- 509–524.

1276 SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 [Link]
Player, A.N., Shen, L.P., Kenny, D., Antao, V.P., and Kolberg, J.A. (2001). Single-copy Reza, K.K., Dey, S., Wuethrich, A., Jing Wang, A., Behren, A., Antaw, F., Wang, Y.,
gene detection using branched DNA (bDNA) in situ hybridization. J Histochem Sina, A.A.I., and Trau, M. (2021). In situ single cell proteomics reveals circulating
Cytochem 49, 603–611. tumor cell heterogeneity during treatment. ACS Nano 15, 11231–11243.
Pliner, H.A., Shendure, J., and Trapnell, C. (2019). Supervised classification enables Ring, A., Nguyen-Sträuli, B.D., Wicki, A., and Aceto, N. (2023). Biology,
rapid annotation of cell atlases. Nat Methods 16, 983–986. vulnerabilities and clinical applications of circulating tumour cells. Nat Rev Cancer
Poirion, O.B., Zhu, X., Ching, T., and Garmire, L. (2016). Single-cell transcriptomics 23, 95–111.
bioinformatics and computational challenges. Front Genet 7, 163. Risso, D., Perraudeau, F., Gribkova, S., Dudoit, S., and Vert, J.P. (2018). A general and
Polański, K., Young, M.D., Miao, Z., Meyer, K.B., Teichmann, S.A., Park, J.E., and flexible method for signal extraction from single-cell RNA-seq data. Nat Commun 9,
Berger, B. (2020). BBKNN: fast batch alignment of single cell transcriptomes. 284.
Bioinformatics 36, 964–965. Rodriguez-Meira, A., Buck, G., Clark, S.A., Povinelli, B.J., Alcolea, V., Louka, E.,
Pott, S. (2017). Simultaneous measurement of chromatin accessibility, DNA McGowan, S., Hamblin, A., Sousos, N., Barkas, N., et al. (2019). Unravelling
methylation, and nucleosome phasing in single cells. eLife 6, e23203. intratumoral heterogeneity through high-sensitivity single-cell mutational analysis
Prasetyanti, P.R., and Medema, J.P. (2017). Intra-tumor heterogeneity from a cancer and parallel RNA sequencing. Mol Cell 73, 1292–1305.e8.
stem cell perspective. Mol Cancer 16, 41. Rodriques, S.G., Stickels, R.R., Goeva, A., Martin, C.A., Murray, E., Vanderburg, C.R.,
Preissl, S., Gaulton, K.J., and Ren, B. (2023). Characterizing cis-regulatory elements Welch, J., Chen, L.M., Chen, F., and Macosko, E.Z. (2019). Slide-seq: a scalable
using single-cell epigenomics. Nat Rev Genet 24, 21–43. technology for measuring genome-wide expression at high spatial resolution.
Qian, J., Liao, J., Liu, Z., Chi, Y., Fang, Y., Zheng, Y., Shao, X., Liu, B., Cui, Y., Guo, W., Science 363, 1463–1467.
et al. (2023). Reconstruction of the cell pseudo-space from single-cell RNA Rood, J.E., Maartens, A., Hupalowska, A., Teichmann, S.A., and Regev, A. (2022).
sequencing data with scSpace. Nat Commun 14, 2484. Impact of the Human Cell Atlas on medicine. Nat Med 28, 2486–2496.
Qiu, C., Martin, B.K., Welsh, I.C., Daza, R.M., Le, T.M., Huang, X., Nichols, E.K., Roohani, Y., Huang, K., and Leskovec, J. (2024). Predicting transcriptional outcomes
Taylor, M.L., Fulton, O., O’Day, D.R., et al. (2024). A single-cell time-lapse of of novel multigene perturbations with GEARS. Nat Biotechnol 42, 927–935.
mouse prenatal development from gastrula to birth. Nature 626, 1084–1093. Rosen, Y., Brbić, M., Roohani, Y., Swanson, K., Li, Z., and Leskovec, J. (2024). Toward
Qiu, Q., Hu, P., Qiu, X., Govek, K.W., Cámara, P.G., and Wu, H. (2020). Massively universal cell embeddings: integrating single-cell RNA-seq datasets across species
parallel and time-resolved RNA sequencing in single cells with scNT-seq. Nat with SATURN. Nat Methods 21, 1492–1500.
Methods 17, 991–1001. Rosen, Y., Roohani, Y., Agarwal, A., Samotorčan, L., Quake, S.R., and Leskovec, J.
Qiu, X., Mao, Q., Tang, Y., Wang, L., Chawla, R., Pliner, H.A., and Trapnell, C. (2023). Universal cell embeddings: a foundation model for cell biology. bioRxiv,
(2017). Reversed graph embedding resolves complex single-cell trajectories. Nat 2023.2011.2028.568918.
Methods 14, 979–982. Rosenberg, A.B., Roco, C.M., Muscat, R.A., Kuchina, A., Sample, P., Yao, Z.,
Qu, J., Yang, F., Zhu, T., Wang, Y., Fang, W., Ding, Y., Zhao, X., Qi, X., Xie, Q., Chen, Graybuck, L.T., Peeler, D.J., Mukherjee, S., Chen, W., et al. (2018). Single-cell
M., et al. (2022). A reference single-cell regulomic and transcriptomic map of profiling of the developing mouse brain and spinal cord with split-pool barcoding.
cynomolgus monkeys. Nat Commun 13, 4069. Science 360, 176–182.
Qu, Y., Liao, Z., Wang, X., Zhang, J., and Liu, C. (2019). EFLDO sensitizes liver cancer Rotem, A., Ram, O., Shoresh, N., Sperling, R.A., Goren, A., Weitz, D.A., and Bernstein,
cells to TNFSF10-induced apoptosis in a p53-dependent manner. Mol Med Report B.E. (2015). Single-cell ChIP-seq reveals cell subpopulations defined by chromatin
19, 3799–3806. state. Nat Biotechnol 33, 1165–1172.
Quon, G., Haider, S., Deshwar, A.G., Cui, A., Boutros, P.C., and Morris, Q. (2013). Rouhanifard, S.H., Mellis, I.A., Dunagin, M., Bayatpour, S., Jiang, C.L., Dardani, I.,
Computational purification of individual tumor gene expression profiles leads to Symmons, O., Emert, B., Torre, E., Cote, A., et al. (2019). ClampFISH detects
significant improvements in prognostic prediction. Genome Med 5, 29. individual nucleic acid molecules using click chemistry-based amplification. Nat
Raj, A., van den Bogaard, P., Rifkin, S.A., van Oudenaarden, A., and Tyagi, S. (2008). Biotechnol 37, 84–89.
Imaging individual mRNA molecules using multiple singly labeled probes. Nat Rousseeuw, P.J. (1987). Silhouettes: a graphical aid to the interpretation and
Methods 5, 877–879. validation of cluster analysis. J Comput Appl Math 20, 53–65.
Ramsköld, D., Luo, S., Wang, Y.C., Li, R., Deng, Q., Faridani, O.R., Daniels, G.A., Rubin, A.J., Parker, K.R., Satpathy, A.T., Qi, Y., Wu, B., Ong, A.J., Mumbach, M.R., Ji,
Khrebtukova, I., Loring, J.F., Laurent, L.C., et al. (2012). Full-length mRNA-Seq A.L., Kim, D.S., Cho, S.W., et al. (2019). Coupled single-cell CRISPR screening and
from single-cell levels of RNA and individual circulating tumor cells. Nat Biotechnol epigenomic profiling reveals causal gene regulatory networks. Cell 176, 361–376.
30, 777–782. e17.
Rang, F.J., de Luca, K.L., de Vries, S.S., Valdes-Quezada, C., Boele, E., Nguyen, P.D., Ruohan, W., Yuwei, Z., Mengbo, W., Xikang, F., Jianping, W., and Shuai Cheng, L.
Guerreiro, I., Sato, Y., Kimura, H., Bakkers, J., et al. (2022). Single-cell profiling of (2022). Resolving single-cell copy number profiling for large datasets. Brief Bioinf
transcriptome and histone modifications with EpiDamID. Mol Cell 82, 1956–1970. 23, bbac264.
e14. Russell, A.J.C., Weir, J.A., Nadaf, N.M., Shabet, M., Kumar, V., Kambhampati, S.,
Rappez, L., Stadler, M., Triana, S., Gathungu, R.M., Ovchinnikova, K., Phapale, P., Raichur, R., Marrero, G.J., Liu, S., Balderrama, K.S., et al. (2024). Slide-tags enables
Heikenwalder, M., and Alexandrov, T. (2021). SpaceM reveals metabolic states of single-nucleus barcoding for multimodal spatial genomics. Nature 625, 101–109.
single cells. Nat Methods 18, 799–805. Ryu, Y., Han, G.H., Jung, E., and Hwang, D. (2023). Integration of single-cell RNA-
Rashid, S., Shah, S., Bar-Joseph, Z., Pandya, R., and Janet, K. (2021). Dhaka: Seq datasets: a review of computational methods. Molecules Cells 46, 106–119.
variational autoencoder for unmasking tumor heterogeneity from single cell Salek, M., Li, N., Chou, H.P., Saini, K., Jovic, A., Jacobs, K.B., Johnson, C., Lu, V., Lee,
genomic data. Bioinformatics 37, 1535–1543. E.J., Chang, C., et al. (2023). COSMOS: a platform for real-time morphology-based,
Redit, C., Cha, S., and Ai, N. (2023). Single-cell proteomics: challenges and prospects. label-free cell sorting using deep learning. Commun Biol 6, 971.
Nature methods 20, 317–318. Salmen, F., De Jonghe, J., Kaminski, T.S., Alemany, A., Parada, G.E., Verity-Legg, J.,
Reed, B.D., Meyer, M.J., Abramzon, V., Ad PhD, O., Ad, O., Adcock, P., Ahmad, F.R., Yanagida, A., Kohler, T.N., Battich, N., van den Brekel, F., et al. (2022). High-
Alppay, G., Ball, J.A., Beach, J., et al. (2022). Real-time dynamic single-molecule throughput total RNA sequencing in single cells using VASA-seq. Nat Biotechnol
protein sequencing on an integrated semiconductor device. Science 378, 186–192. 40, 1780–1793.
Regev, A., Teichmann, S., Rozenblatt-Rosen, O., Stubbington, M., Ardlie, K., Amit, I., Sampath Kumar, A., Tian, L., Bolondi, A., Hernández, A.A., Stickels, R., Kretzmer, H.,
Arlotta, P., Bader, G., Benoist, C., and Biton, M. (2018). The human cell atlas white Murray, E., Wittler, L., Walther, M., Barakat, G., et al. (2023). Spatiotemporal
paper. arXiv,181005192. transcriptomic maps of whole mouse embryos at the onset of organogenesis. Nat
Regev, A., Teichmann, S.A., Lander, E.S., Amit, I., Benoist, C., Birney, E., Bodenmiller, Genet 55, 1176–1185.
B., Campbell, P., Carninci, P., Clatworthy, M., et al. (2017). The Human Cell Atlas. Sankaran, V.G., Weissman, J.S., and Zon, L.I. (2022). Cellular barcoding to decipher
eLife 6, e27041. clonal dynamics in disease. Science 378, eabm5874.
Reimegård, J., Tarbier, M., Danielsson, M., Schuster, J., Baskaran, S., Panagiotou, S., Sasagawa, Y., Nikaido, I., Hayashi, T., Danno, H., Uno, K.D., Imai, T., and Ueda, H.R.
Dahl, N., Friedländer, M.R., and Gallant, C.J. (2021). A combined approach for (2013). Quartz-Seq: a highly reproducible and sensitive single-cell RNA sequencing
single-cell mRNA and intracellular protein expression analysis. Commun Biol 4, method, reveals non-genetic gene-expression heterogeneity. Genome Biol 14, R31.
624. Sashittal, P., Zhang, H., Iacobuzio-Donahue, C.A., and Raphael, B.J. (2023). ConDoR:
Ren, X., Wen, W., Fan, X., Hou, W., Su, B., Cai, P., Li, J., Liu, Y., Tang, F., Zhang, F., et tumor phylogeny inference with a copy-number constrained mutation loss model.
al. (2021). COVID-19 immune features revealed by a large-scale single-cell Genome Biol 24, 272.
transcriptome atlas. Cell 184, 1895–1913.e19. Sasse, A., Ng, B., Spiro, A.E., Tasaki, S., Bennett, D.A., Gaiteri, C., De Jager, P.L.,
Reuter, J.A., Spacek, D.V., and Snyder, M.P. (2015). High-throughput sequencing Chikina, M., and Mostafavi, S. (2023). Benchmarking of deep neural networks for
technologies. Mol Cell 58, 586–597. predicting personal gene expression from DNA sequence highlights shortcomings.

[Link] SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 1277
 Nat Genet 55, 2060–2064. 820.
Satpathy, A.T., Granja, J.M., Yost, K.E., Qi, Y., Meschi, F., McDermott, G.P., Olsen, B. Smith, R.J., Zhang, H., Hu, S.S., Yung, T., Francis, R., Lee, L., Onaitis, M.W., Dirks, P.
N., Mumbach, M.R., Pierce, S.E., Corces, M.R., et al. (2019). Massively parallel B., Zang, C., and Kim, T.H. (2022). Single-cell chromatin profiling of the primitive
single-cell chromatin landscapes of human immune cell development and gut tube reveals regulatory dynamics underlying lineage fate decisions. Nat
intratumoral T cell exhaustion. Nat Biotechnol 37, 925–936. Commun 13, 2965.
Schede, H.H., Schneider, C.G., Stergiadou, J., Borm, L.E., Ranjak, A., Yamawaki, T.M., Snyder, M.P., Lin, S., Posgai, A., Atkinson, M., Regev, A., Rood, J., Rozenblatt-Rosen,
David, F.P.A., Lönnerberg, P., Tosches, M.A., Codeluppi, S., et al. (2021). Spatial O., Gaffney, L., Hupalowska, A., Satija, R., et al. (2019). The human body at
tissue profiling by imaging-free molecular tomography. Nat Biotechnol 39, 968– cellular resolution: the NIH Human Biomolecular Atlas Program. Nature 574,
977. 187–192.
Schott, M., León-Periñán, D., Splendiani, E., Strenger, L., Licha, J.R., Pentimalli, T.M., Sokolova, K., Theesfeld, C.L., Wong, A.K., Zhang, Z., Dolinski, K., and Troyanskaya, O.
Schallenberg, S., Alles, J., Samut Tagliaferro, S., Boltengagen, A., et al. (2024). G. (2023). Atlas of primary cell-type-specific sequence models of gene expression
Open-ST: high-resolution spatial transcriptomics in 3D. Cell 187, 3953–3972.e26. and variant effects. Cell Rep Methods 3, 100580.
Schreibing, F., and Kramann, R. (2022). Mapping the human kidney using single-cell Song, L., Zhang, Z., Grasfeder, L.L., Boyle, A.P., Giresi, P.G., Lee, B.K., Sheffield, N.C.,
genomics. Nat Rev Nephrol 18, 347–360. Gräf, S., Huss, M., Keefe, D., et al. (2011). Open chromatin defined by DNaseI and
Schulz, D., Zanotelli, V.R.T., Fischer, J.R., Schapiro, D., Engler, S., Lun, X.K., Jackson, FAIRE identifies regulatory elements that shape cell-type identity. Genome Res 21,
H.W., and Bodenmiller, B. (2018). Simultaneous multiplexed imaging of mRNA 1757–1767.
and proteins with subcellular resolution in breast cancer tissue samples by mass Song, Q., and Su, J. (2021). DSTG: Deconvoluting spatial transcriptomics data
cytometry. Cell Syst 6, 25–36.e5. through graph-based artificial intelligence. Brief Bioinf 22, bbaa414.
Schwessinger, R., Deasy, J., Woodruff, R.T., Young, S., and Branson, K.M. (2023). Song, Q., Su, J., and Zhang, W. (2021). scGCN is a graph convolutional networks
Single-cell gene expression prediction from DNA sequence at large contexts. algorithm for knowledge transfer in single cell omics. Nat Commun 12, 3826.
bioRxiv, 2023.2007.2026.550634. Song, T., Dai, H., Wang, S., Wang, G., Zhang, X., Zhang, Y., and Jiao, L. (2022).
Sender, R., Weiss, Y., Navon, Y., Milo, I., Azulay, N., Keren, L., Fuchs, S., Ben-Zvi, D., TransCluster: a cell-type identification method for single-cell RNA-Seq data using
Noor, E., and Milo, R. (2023). The total mass, number, and distribution of immune deep learning based on transformer. Front Genet 13, 1038919.
cells in the human body. Proc Natl Acad Sci USA 120, e2308511120. Song, Y., Miao, Z., Brazma, A., and Papatheodorou, I. (2023). Benchmarking
Seninge, L., Anastopoulos, I., Ding, H., and Stuart, J. (2021). VEGA is an interpretable strategies for cross-species integration of single-cell RNA sequencing data. Nat
generative model for inferring biological network activity in single-cell transcrip- Commun 14, 6495.
tomics. Nat Commun 12, 5684. Sos, B.C., Fung, H.L., Gao, D.R., Osothprarop, T.F., Kia, A., He, M.M., and Zhang, K.
Shah, S., Lubeck, E., Schwarzkopf, M., He, T.F., Greenbaum, A., Sohn, C.H., Lignell, (2016). Characterization of chromatin accessibility with a transposome hypersen-
A., Choi, H.M.T., Gradinaru, V., Pierce, N.A., et al. (2016). Single-molecule RNA sitive sites sequencing (THS-seq) assay. Genome Biol 17, 20.
detection at depth by hybridization chain reaction and tissue hydrogel embedding Srivatsan, S.R., McFaline-Figueroa, J.L., Ramani, V., Saunders, L., Cao, J., Packer, J.,
and clearing. Development 143, 2862–2867. Pliner, H.A., Jackson, D.L., Daza, R.M., Christiansen, L., et al. (2020). Massively
Shalek, A.K., Satija, R., Shuga, J., Trombetta, J.J., Gennert, D., Lu, D., Chen, P., multiplex chemical transcriptomics at single-cell resolution. Science 367, 45–51.
Gertner, R.S., Gaublomme, J.T., Yosef, N., et al. (2014). Single-cell RNA-seq reveals Ståhl, P.L., Salmén, F., Vickovic, S., Lundmark, A., Navarro, J.F., Magnusson, J.,
dynamic paracrine control of cellular variation. Nature 510, 363–369. Giacomello, S., Asp, M., Westholm, J.O., Huss, M., et al. (2016). Visualization and
Shao, Y., Wang, X.B., Zhang, J.J., Li, M.L., Wu, S.S., Ma, X.Y., Wang, X., Zhao, H.F., analysis of gene expression in tissue sections by spatial transcriptomics. Science
Li, Y., Zhu, H.H., et al. (2020). Genome and single-cell RNA-sequencing of the 353, 78–82.
earthworm Eisenia andrei identifies cellular mechanisms underlying regeneration. Stewart, B.J., Ferdinand, J.R., and Clatworthy, M.R. (2020). Using single-cell
Nat Commun 11, 2656. technologies to map the human immune system—implications for nephrology.
Sharma, P., Goswami, S., Raychaudhuri, D., Siddiqui, B.A., Singh, P., Nagarajan, A., Nat Rev Nephrol 16, 112–128.
Liu, J., Subudhi, S.K., Poon, C., Gant, K.L., et al. (2023). Immune checkpoint Stickels, R.R., Murray, E., Kumar, P., Li, J., Marshall, J.L., Di Bella, D.J., Arlotta, P.,
therapy—current perspectives and future directions. Cell 186, 1652–1669. Macosko, E.Z., and Chen, F. (2021). Highly sensitive spatial transcriptomics at
Shekhar, K., Lapan, S.W., Whitney, I.E., Tran, N.M., Macosko, E.Z., Kowalczyk, M., near-cellular resolution with Slide-seqV2. Nat Biotechnol 39, 313–319.
Adiconis, X., Levin, J.Z., Nemesh, J., Goldman, M., et al. (2016). Comprehensive Stoeckius, M., Hafemeister, C., Stephenson, W., Houck-Loomis, B., Chattopadhyay, P.
classification of retinal bipolar neurons by single-cell transcriptomics. Cell 166, K., Swerdlow, H., Satija, R., and Smibert, P. (2017). Simultaneous epitope and
1308–1323.e30. transcriptome measurement in single cells. Nat Methods 14, 865–868.
Shen-Orr, S.S., Tibshirani, R., Khatri, P., Bodian, D.L., Staedtler, F., Perry, N.M., Stoeckius, M., Zheng, S., Houck-Loomis, B., Hao, S., Yeung, B.Z., Mauck Iii, W.M.,
Hastie, T., Sarwal, M.M., Davis, M.M., and Butte, A.J. (2010). Cell type-specific gene Smibert, P., and Satija, R. (2018). Cell Hashing with barcoded antibodies enables
expression differences in complex tissues. Nat Methods 7, 287–289. multiplexing and doublet detection for single cell genomics. Genome Biol 19, 224.
Shen, H., Liu, J., Hu, J., Shen, X., Zhang, C., Wu, D., Feng, M., Yang, M., Li, Y., Yang, Street, K., Risso, D., Fletcher, R.B., Das, D., Ngai, J., Yosef, N., Purdom, E., and Dudoit,
Y., et al. (2023). Generative pretraining from large-scale transcriptomes for single- S. (2018). Slingshot: Cell lineage and pseudotime inference for single-cell
cell deciphering. iScience 26, 106536. transcriptomics. BMC Genomics 19, 477.
Shen, H., Shen, X., Feng, M., Wu, D., Zhang, C., Yang, Y., Yang, M., Hu, J., Liu, J., Stuart, T., Hao, S., Zhang, B., Mekerishvili, L., Landau, D.A., Maniatis, S., Satija, R.,
Wang, W., et al. (2022). A universal approach for integrating super large-scale and Raimondi, I. (2022). Nanobody-tethered transposition allows for multifactorial
single-cell transcriptomes by exploring gene rankings. Brief Bioinf 23, bbab573. chromatin profiling at single-cell resolution. bioRxiv, 2022.03.08.483436.
Sheng, K., Cao, W., Niu, Y., Deng, Q., and Zong, C. (2017). Effective detection of Stuart, T., Srivastava, A., Madad, S., Lareau, C.A., and Satija, R. (2021). Single-cell
variation in single-cell transcriptomes using MATQ-seq. Nat Methods 14, 267–270. chromatin state analysis with Signac. Nat Methods 18, 1333–1341.
Shi, H., He, Y., Zhou, Y., Huang, J., Maher, K., Wang, B., Tang, Z., Luo, S., Tan, P., Stubbington, M.J.T., Rozenblatt-Rosen, O., Regev, A., and Teichmann, S.A. (2017).
Wu, M., et al. (2023). Spatial atlas of the mouse central nervous system at Single-cell transcriptomics to explore the immune system in health and disease.
molecular resolution. Nature 622, 552–561. Science 358, 58–63.
Shrestha, B. (2020). Single-Cell Metabolomics by Mass Spectrometry. Methods Mol Sui, X., Lo, J.A., Luo, S., He, Y., Tang, Z., Lin, Z., Zhou, Y., Wang, W.X., Liu, J., and
Biol 2064, 1–8. Wang, X. (2024). Scalable spatial single-cell transcriptomics and translatomics in
Sibai, M., Altuntaş, E., Süzek, B.ı.E., Şahin, B., Parlayan, C., Öztürk, G., Baykal, A.T.ı., 3D thick tissue blocks. bioRxiv, 2024.2008.2005.606553.
and Demircan, T. (2020). Comparison of protein expression profile of limb Sun, T., Wu, R., and Ming, L. (2019). The role of m6A RNA methylation in cancer.
regeneration between neotenic and metamorphic axolotl. Biochem Biophys Res Biomed Pharmacother 112, 108613.
Commun 522, 428–434. Sun, Z., Tang, Y., Zhang, Y., Fang, Y., Jia, J., Zeng, W., and Fang, D. (2021a). Joint
Sikkema, L., Ramírez-Suástegui, C., Strobl, D.C., Gillett, T.E., Zappia, L., Madissoon, E., single-cell multiomic analysis in Wnt3a induced asymmetric stem cell division. Nat
Markov, N.S., Zaragosi, L.E., Ji, Y., Ansari, M., et al. (2023). An integrated cell atlas Commun 12, 5941.
of the lung in health and disease. Nat Med 29, 1563–1577. Sun, Z., Vaisvila, R., Hussong, L.M., Yan, B., Baum, C., Saleh, L., Samaranayake, M.,
Siletti, K., Hodge, R., Albiach, A.M., Hu, L., Lee, K.W., Lönnerberg, P., Bakken, T., Guan, S., Dai, N., Corrêa Jr., I.R., et al. (2021b). Nondestructive enzymatic
Ding, S.L., Clark, M., Casper, T., et al. (2022). Transcriptomic diversity of cell types deamination enables single-molecule long-read amplicon sequencing for the
across the adult human brain. bioRxiv, 2022.2010.2012.511898. determination of 5-methylcytosine and 5-hydroxymethylcytosine at single-base
Smallwood, S.A., Lee, H.J., Angermueller, C., Krueger, F., Saadeh, H., Peat, J., resolution. Genome Res 31, 291–300.
Andrews, S.R., Stegle, O., Reik, W., and Kelsey, G. (2014). Single-cell genome-wide Suo, C., Dann, E., Goh, I., Jardine, L., Kleshchevnikov, V., Park, J.E., Botting, R.A.,
bisulfite sequencing for assessing epigenetic heterogeneity. Nat Methods 11, 817– Stephenson, E., Engelbert, J., Tuong, Z.K., et al. (2022). Mapping the developing

1278 SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 [Link]
 human immune system across organs. Science 376, eabo0510. S. (2021). Single-cell transcriptomic characterization of a gastrulating human
Suvilesh, K.N., Nussbaum, Y.I., Radhakrishnan, V., Manjunath, Y., Avella, D.M., embryo. Nature 600, 285–289.
Staveley-O’Carroll, K.F., Kimchi, E.T., Chaudhuri, A.A., Shyu, C.R., Li, G., et al. Vallejo, A.F., Harvey, K., Wang, T., Wise, K., Butler, L.M., Polo, J., Plummer, J.,
(2022). Tumorigenic circulating tumor cells from xenograft mouse models of non- Swarbrick, A., and Martelotto, L.G. (2022). snPATHO-seq: unlocking the FFPE
metastatic NSCLC patients reveal distinct single cell heterogeneity and drug archives for single nucleus RNA profiling. bioRxiv, 2022.2008.2023.505054.
responses. Mol Cancer 21, 73. Van de Sande, B., Lee, J.S., Mutasa-Gottgens, E., Naughton, B., Bacon, W., Manning,
Svensson, V., da-Veiga-Beltrame, E., and Pachter, L. (2020). A curated database J., Wang, Y., Pollard, J., Mendez, M., Hill, J., et al. (2023). Applications of single-cell
reveals trends in single-cell transcriptomics. Database 2020, baaa073. RNA sequencing in drug discovery and development. Nat Rev Drug Discov 22,
Tabula Muris, C., Overall, c., Logistical, c., Organ, c., processing, Library, p., 496–520.
sequencing, Computational data, a., Cell type, a., Writing, g., et al. (2018). Single- van Dijk, D., Sharma, R., Nainys, J., Yim, K., Kathail, P., Carr, A.J., Burdziak, C.,
cell transcriptomics of 20 mouse organs creates a Tabula Muris. Nature 562, 367– Moon, K.R., Chaffer, C.L., Pattabiraman, D., et al. (2018). Recovering gene
372. interactions from single-cell data using data diffusion. Cell 174, 716–729.e27.
Takei, Y., Yun, J., Zheng, S., Ollikainen, N., Pierson, N., White, J., Shah, S., Thomassie, Vandereyken, K., Sifrim, A., Thienpont, B., and Voet, T. (2023). Methods and
J., Suo, S., Eng, C.H.L., et al. (2021). Integrated spatial genomics reveals global applications for single-cell and spatial multi-omics. Nat Rev Genet 24, 494–515.
architecture of single nuclei. Nature 590, 344–350. VanInsberghe, M., van den Berg, J., Andersson-Rolf, A., Clevers, H., and van
Takeshima, H., and Ushijima, T. (2019). Accumulation of genetic and epigenetic Oudenaarden, A. (2021). Single-cell Ribo-seq reveals cell cycle-dependent
alterations in normal cells and cancer risk. npj Precis Onc 3, 7. translational pausing. Nature 597, 561–565.
Tan, L., Ma, W., Wu, H., Zheng, Y., Xing, D., Chen, R., Li, X., Daley, N., Deisseroth, K., Varrone, M., Tavernari, D., Santamaria-Martínez, A., Walsh, L.A., and Ciriello, G.
and Xie, X.S. (2021). Changes in genome architecture and transcriptional (2024). CellCharter reveals spatial cell niches associated with tissue remodeling and
dynamics progress independently of sensory experience during post-natal brain cell plasticity. Nat Genet 56, 74–84.
development. Cell 184, 741–758.e17. Velmeshev, D., Perez, Y., Yan, Z., Valencia, J.E., Castaneda-Castellanos, D.R., Wang,
Tan, L., Xing, D., Chang, C.H., Li, H., and Xie, X.S. (2018). Three-dimensional genome L., Schirmer, L., Mayer, S., Wick, B., Wang, S., et al. (2023). Single-cell analysis of
structures of single diploid human cells. Science 361, 924–928. prenatal and postnatal human cortical development. Science 382, eadf0834.
Tan, S.J., Li, Q., and Lim, C.T. (2010). Manipulation and isolation of single cells and Vento-Tormo, R., Efremova, M., Botting, R.A., Turco, M.Y., Vento-Tormo, M., Meyer,
nuclei. Methods Cell Biol 98, 79–96. K.B., Park, J.E., Stephenson, E., POLAŃSKI, K., Goncalves, A., et al. (2018). Single-
Tanay, A., and Regev, A. (2017). Scaling single-cell genomics from phenomenology to cell reconstruction of the early maternal-fetal interface in humans. Nature 563,
mechanism. Nature 541, 331–338. 347–353.
Tang, F., Barbacioru, C., Wang, Y., Nordman, E., Lee, C., Xu, N., Wang, X., Bodeau, J., Vickovic, S., Eraslan, G., Salmén, F., Klughammer, J., Stenbeck, L., Schapiro, D., Äijö,
Tuch, B.B., Siddiqui, A., et al. (2009). mRNA-Seq whole-transcriptome analysis of a T., Bonneau, R., Bergenstråhle, L., Navarro, J.F., et al. (2019). High-definition
single cell. Nat Methods 6, 377–382. spatial transcriptomics for in situ tissue profiling. Nat Methods 16, 987–990.
Tang, F., Li, J., Qi, L., Liu, D., Bo, Y., Qin, S., Miao, Y., Yu, K., Hou, W., Li, J., et al. Vu, T.N., Nguyen, H.N., Calza, S., Kalari, K.R., Wang, L., Pawitan, Y., and Berger, B.
(2023a). A pan-cancer single-cell panorama of human natural killer cells. Cell 186, (2019). Cell-level somatic mutation detection from single-cell RNA sequencing.
4235–4251.e20. Bioinformatics 35, 4679–4687.
Tang, L. (2021). Multiomics sequencing goes spatial. Nat Methods 18, 31. Wagner, A., Wang, C., Fessler, J., DeTomaso, D., Avila-Pacheco, J., Kaminski, J.,
Tang, W., Bertaux, F., Thomas, P., Stefanelli, C., Saint, M., Marguerat, S., Shahrezaei, Zaghouani, S., Christian, E., Thakore, P., Schellhaass, B., et al. (2021). Metabolic
V., and Kelso, J. (2020). bayNorm: Bayesian gene expression recovery, imputation modeling of single Th17 cells reveals regulators of autoimmunity. Cell 184, 4168–
and normalization for single-cell RNA-sequencing data. Bioinformatics 36, 1174– 4185.e21.
1181. Wagner, F., Yan, Y., and Yanai, I. (2018). K-nearest neighbor smoothing for high-
Tang, W., Wen, H., Liu, R., Ding, J., Jin, W., Xie, Y., Liu, H., and Tang, J. (2023b). throughput single-cell RNA-Seq data. bioRxiv, 217737.
Single-Cell Multimodal Prediction via Transformers. arXiv, 2303.00233. Walker, A., and Parkhill, J. (2008). Single-cell genomics. Nat Rev Microbiol 6, 176–
Tang, X., Chen, J., Zhang, X., Liu, X., Xie, Z., Wei, K., Qiu, J., Ma, W., Lin, C., and Ke, 177.
R. (2023c). Improved in situ sequencing for high-resolution targeted spatial Wan, X., Xiao, J., Tam, S.S.T., Cai, M., Sugimura, R., Wang, Y., Wan, X., Lin, Z., Wu,
transcriptomic analysis in tissue sections. J Genet Genomics 50, 652–660. A.R., and Yang, C. (2023). Integrating spatial and single-cell transcriptomics data
Tarashansky, A.J., Musser, J.M., Khariton, M., Li, P., Arendt, D., Quake, S.R., and using deep generative models with SpatialScope. Nat Commun 14, 7848.
Wang, B. (2021). Mapping single-cell atlases throughout Metazoa unravels cell Wang, B., Lin, A.E., Yuan, J., Novak, K.E., Koch, M.D., Wingreen, N.S., Adamson, B.,
type evolution. eLife 10, 10.7554–66747. and Gitai, Z. (2023a). Single-cell massively-parallel multiplexed microbial sequen-
Tedesco, M., Giannese, F., Lazarević, D., Giansanti, V., Rosano, D., Monzani, S., cing (M3-seq) identifies rare bacterial populations and profiles phage infection. Nat
Catalano, I., Grassi, E., Zanella, E.R., Botrugno, O.A., et al. (2022). Chromatin Microbiol 8, 1846–1862.
velocity reveals epigenetic dynamics by single-cell profiling of heterochromatin and Wang, F., Wang, Q., Mohanty, V., Liang, S., Dou, J., Han, J., Minussi, D.C., Gao, R.,
euchromatin. Nat Biotechnol 40, 235–244. Ding, L., Navin, N., et al. (2021a). MEDALT: Single-cell copy number lineage
The Tabula Sapiens Consortium, Jones, R.C., Karkanias, J., Krasnow, M.A., Pisco, A. tracing enabling gene discovery. Genome Biol 22, 70.
O., Quake, S.R., Salzman, J., Yosef, N., Bulthaup, B., Brown, P., et al. (2022). The Wang, G., Moffitt, J.R., and Zhuang, X. (2018a). Multiplexed imaging of high-density
Tabula Sapiens: a multiple-organ, single-cell transcriptomic atlas of humans. libraries of RNAs with MERFISH and expansion microscopy. Sci Rep 8, 4847.
Science 376, eabl4896. Wang, J., Chen, L., Chen, Z., and Zhang, W. (2015). RNA-seq based transcriptomic
Theodoris, C.V., Xiao, L., Chopra, A., Chaffin, M.D., Al Sayed, Z.R., Hill, M.C., analysis of single bacterial cells. Integr Biol 7, 1466–1476.
Mantineo, H., Brydon, E.M., Zeng, Z., Liu, X.S., et al. (2023). Transfer learning Wang, J., Sun, H., Jiang, M., Li, J., Zhang, P., Chen, H., Mei, Y., Fei, L., Lai, S., Han, X.,
enables predictions in network biology. Nature 618, 616–624. et al. (2021b). Tracing cell-type evolution by cross-species comparison of cell
Tian, F., Zhou, F., Li, X., Ma, W., Wu, H., Yang, M., Chapman, A.R., Lee, D.F., Tan, L., atlases. Cell Rep 34, 108803.
Xing, D., et al. (2020). Genomic architecture of cells in tissues (GeACT): study of Wang, Q., Xiong, H., Ai, S., Yu, X., Liu, Y., Zhang, J., and He, A. (2019a). CoBATCH
human mid-gestation fetus. bioRxiv, 2020.2004.2012.038000. for high-throughput single-cell epigenomic profiling. Mol Cell 76, 206–216.e7.
Tian, L., Chen, F., and Macosko, E.Z. (2023a). The expanding vistas of spatial Wang, R., and Jiang, Y. (2022). Copy number variation detection by single-cell DNA
transcriptomics. Nat Biotechnol 41, 773–782. sequencing with SCOPE. Methods Mol Biol 2493, 279– 288.
Tian, W., Zhou, J., Bartlett, A., Zeng, Q., Liu, H., Castanon, R.G., Kenworthy, M., Wang, R., Lin, D.Y., and Jiang, Y. (2020a). SCOPE: a normalization and copy-number
Altshul, J., Valadon, C., Aldridge, A., et al. (2023b). Single-cell DNA methylation estimation method for single-cell DNA sequencing. Cell Syst 10, 445–452.e6.
and 3D genome architecture in the human brain. Science 382, eadf5357. Wang, R., Zhang, P., Wang, J., Ma, L., E, W., Suo, S., Jiang, M., Li, J., Chen, H., Sun,
Tran, D., Nguyen, H., Tran, B., La Vecchia, C., Luu, H.N., and Nguyen, T. (2021). Fast H., et al. (2023b). Construction of a cross-species cell landscape at single-cell level.
and precise single-cell data analysis using a hierarchical autoencoder. Nat Nucleic Acids Res 51, 501–516.
Commun 12, 1029. Wang, S., Zheng, Y., Li, J., Yu, Y., Zhang, W., Song, M., Liu, Z., Min, Z., Hu, H., Jing,
Tran, H.T.N., Ang, K.S., Chevrier, M., Zhang, X., Lee, N.Y.S., Goh, M., and Chen, J. Y., et al. (2020b). Single-cell transcriptomic atlas of primate ovarian aging. Cell
(2020). A benchmark of batch-effect correction methods for single-cell RNA 180, 585–600.e19.
sequencing data. Genome Biol 21, 12. Wang, X., Allen, W.E., Wright, M.A., Sylwestrak, E.L., Samusik, N., Vesuna, S., Evans,
Tsompana, M., and Buck, M.J. (2014). Chromatin accessibility: a window into the K., Liu, C., Ramakrishnan, C., Liu, J., et al. (2018b). Three-dimensional intact-
genome. Epigenet Chromatin 7, 33. tissue sequencing of single-cell transcriptional states. Science 361, eaat5691.
Tyser, R.C.V., Mahammadov, E., Nakanoh, S., Vallier, L., Scialdone, A., and Srinivas, Wang, X., Park, J., Susztak, K., Zhang, N.R., and Li, M. (2019b). Bulk tissue cell type

[Link] SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 1279
 deconvolution with multi-subject single-cell expression reference. Nat Commun 10, J., Dröge, P., Koh, C.G., et al. (2020). Parallel bimodal single-cell sequencing of
380. transcriptome and chromatin accessibility. Genome Res 30, 1027–1039.
Wang, Y., Eddison, M., Fleishman, G., Weigert, M., Xu, S., Wang, T., Rokicki, K., Xiong, H., Luo, Y., Wang, Q., Yu, X., and He, A. (2021). Single-cell joint detection of
Goina, C., Henry, F.E., Lemire, A.L., et al. (2021c). EASI-FISH for thick tissue chromatin occupancy and transcriptome enables higher-dimensional epigenomic
defines lateral hypothalamus spatio-molecular organization. Cell 184, 6361–6377. reconstructions. Nat Methods 18, 652–660.
e24. Xiong, L., Chen, T., and Kellis, M. (2023). scCLIP: multi-modal single-cell contrastive
Wang, Y., Fan, J.L., Melms, J.C., Amin, A.D., Georgis, Y., Barrera, I., Ho, P., Tagore, S., learning integration pre-training. In: NeurIPS 2023 AI for Science Workshop.
Abril-Rodríguez, G., He, S., et al. (2023c). Multimodal single-cell and whole- Available from URL: [Link]
genome sequencing of small, frozen clinical specimens. Nat Genet 55, 19–25. Xiong, L., Xu, K., Tian, K., Shao, Y., Tang, L., Gao, G., Zhang, M., Jiang, T., and
Wang, Y., Yuan, P., Yan, Z., Yang, M., Huo, Y., Nie, Y., Zhu, X., Qiao, J., and Yan, L. Zhang, Q.C. (2019). SCALE method for single-cell ATAC-seq analysis via latent
(2021d). Single-cell multiomics sequencing reveals the functional regulatory feature extraction. Nat Commun 10, 4576.
landscape of early embryos. Nat Commun 12, 1247. Xiong, Z., Xu, K., Lin, Z., Kong, F., Wang, Q., Quan, Y., Sha, Q., Li, F., Zou, Z., Liu, L.,
Weber, E.W., Maus, M.V., and Mackall, C.L. (2020). The emerging landscape of et al. (2022). Ultrasensitive Ribo-seq reveals translational landscapes during
immune cell therapies. Cell 181, 46–62. mammalian oocyte-to-embryo transition and pre-implantation development. Nat
Weber, L.L., Sashittal, P., and El-Kebir, M. (2021). doubletD: Detecting doublets in Cell Biol 24, 968–980.
single-cell DNA sequencing data. Bioinformatics 37, i214–i221. Xu, C., Jin, X., Wei, S., Wang, P., Luo, M., Xu, Z., Yang, W., Cai, Y., Xiao, L., Lin, X., et
Weber, L.L., Zhang, C., Ochoa, I., El-Kebir, M., and Przytycka, T.M. (2023). Phertilizer: al. (2022a). DeepST: Identifying spatial domains in spatial transcriptomics by deep
growing a clonal tree from ultra-low coverage single-cell DNA sequencing of learning. Nucleic Acids Res 50, e131.
tumors. PLoS Comput Biol 19, e1011544. Xu, C., Lopez, R., Mehlman, E., Regier, J., Jordan, M.I., and Yosef, N. (2021a).
Wei, X., Fu, S., Li, H., Liu, Y., Wang, S., Feng, W., Yang, Y., Liu, X., Zeng, Y.Y., Cheng, Probabilistic harmonization and annotation of single-cell transcriptomics data with
M., et al. (2022). Single-cell Stereo-seq reveals induced progenitor cells involved in deep generative models. Mol Syst Biol 17, e9620.
axolotl brain regeneration. Science 377, eabp9444. Xu, C., Prete, M., Webb, S., Jardine, L., Stewart, B.J., Hoo, R., He, P., Meyer, K.B., and
Weinstein, J.A., Regev, A., and Zhang, F. (2019). DNA microscopy: optics-free spatio- Teichmann, S.A. (2023a). Automatic cell-type harmonization and integration
genetic imaging by a stand-alone chemical reaction. Cell 178, 229–241.e16. across Human Cell Atlas datasets. Cell 186, 5876–5891.e20.
Welch, J., Kozareva, V., Ferreira, A., Vanderburg, C., Martin, C., and Macosko, E. Xu, J., Zhang, A., Liu, F., Chen, L., and Zhang, X. (2023b). CIForm as a Transformer-
(2018). Integrative inference of brain cell similarities and differences from single- based model for cell-type annotation of large-scale single-cell RNA-seq data. Brief
cell genomics. bioRxiv, 459891. Bioinf 24, bbad195.
Wen, L., and Tang, F. (2022). Recent advances in single-cell sequencing technologies. Xu, J., Zhang, A., Liu, F., Zhang, X., and Birol, I. (2023c). STGRNS: An interpretable
Precision Clin Med 5, pbac002. transformer-based method for inferring gene regulatory networks from single-cell
Wen, N., Zhao, Z., Fan, B., Chen, D., Men, D., Wang, J., and Chen, J. (2016). transcriptomic data. Bioinformatics 39, btad165.
Development of droplet microfluidics enabling high-throughput single-cell analysis. Xu, W., Wen, Y., Liang, Y., Xu, Q., Wang, X., Jin, W., and Chen, X. (2021b). A plate-
Molecules 21, 881. based single-cell ATAC-seq workflow for fast and robust profiling of chromatin
Weninger, A., and Arlotta, P. (2023). A family portrait of human brain cells. Science accessibility. Nat Protoc 16, 4084–4107.
382, 168–169. Xu, W., Yang, W., Zhang, Y., Chen, Y., Hong, N., Zhang, Q., Wang, X., Hu, Y., Song,
Westermann, A.J., and Vogel, J. (2021). Cross-species RNA-seq for deciphering host- K., Jin, W., et al. (2022b). ISSAAC-seq enables sensitive and flexible multimodal
microbe interactions. Nat Rev Genet 22, 361–378. profiling of chromatin accessibility and gene expression in single cells. Nat Methods
Wherry, E.J. (2011). T cell exhaustion. Nat Immunol 12, 492–499. 19, 1243–1249.
Winkler, E.A., Kim, C.N., Ross, J.M., Garcia, J.H., Gil, E., Oh, I., Chen, L.Q., Wu, D., Xu, Y., Zhang, T., Zhou, Q., Hu, M., Qi, Y., Xue, Y., Nie, Y., Wang, L., Bao, Z., and Shi,
Catapano, J.S., Raygor, K., et al. (2022). A single-cell atlas of the normal and W. (2023d). A single-cell transcriptome atlas profiles early organogenesis in
malformed human brain vasculature. Science 375, eabi7377. human embryos. Nat Cell Biol 25, 604–615.
Winkler, I., Tolkachov, A., Lammers, F., Lacour, P., Daugelaite, K., Schneider, N., Xu, Z., Wang, Y., Sheng, K., Rosenthal, R., Liu, N., Hua, X., Zhang, T., Chen, J., Song,
Koch, M.L., Panten, J., Grünschläger, F., Poth, T., et al. (2024). The cycling and M., Lv, Y., et al. (2023e). Droplet-based high-throughput single microbe RNA
aging mouse female reproductive tract at single-cell resolution. Cell 187, 981–998. sequencing by smRandom-seq. Nat Commun 14, 5130.
e25. Xu, Z., Zhang, T., Chen, H., Zhu, Y., Lv, Y., Zhang, S., Chen, J., Chen, H., Yang, L.,
Wolf, F.A., Angerer, P., and Theis, F.J. (2018). SCANPY: large-scale single-cell gene Jiang, W., et al. (2023f). High-throughput single nucleus total RNA sequencing of
expression data analysis. Genome Biol 19, 15. formalin-fixed paraffin-embedded tissues by snRandom-seq. Nat Commun 14,
Wood, L.D., Parsons, D.W., Jones, S., Lin, J., Sjöblom, T., Leary, R.J., Shen, D., Boca, S. 2734.
M., Barber, T., Ptak, J., et al. (2007). The genomic landscapes of human breast and Xue, R., Zhang, Q., Cao, Q., Kong, R., Xiang, X., Liu, H., Feng, M., Wang, F., Cheng, J.,
colorectal cancers. Science 318, 1108–1113. Li, Z., et al. (2022). Liver tumour immune microenvironment subtypes and
Wu, C., Simonetti, M., Rossell, C., Mignardi, M., Mirzazadeh, R., Annaratone, L., neutrophil heterogeneity. Nature 612, 141–147.
Marchiò, C., Sapino, A., Bienko, M., Crosetto, N., et al. (2018). RollFISH achieves Yamada, S., Nagafuchi, Y., Wang, M., Ota, M., Hatano, H., Takeshima, Y., Okubo, M.,
robust quantification of single-molecule RNA biomarkers in paraffin-embedded Kobayashi, S., Sugimori, Y., Masahiro, N., et al. (2023). Immunomics analysis of
tumor tissue samples. Commun Biol 1, 209. rheumatoid arthritis identified precursor dendritic cells as a key cell subset of
Wu, H., Zhang, J., Jian, F., Chen, J.P., Zheng, Y., Tan, L., and Sunney Xie, X. (2024a). treatment resistance. Ann Rheum Dis 82, 809–819.
Simultaneous single-cell three-dimensional genome and gene expression profiling Yamagishi, M., Kuze, Y., Kobayashi, S., Nakashima, M., Morishima, S., Kawamata, T.,
uncovers dynamic enhancer connectivity underlying olfactory receptor choice. Nat Makiyama, J., Suzuki, K., Seki, M., Abe, K., et al. (2024). Mechanisms of action and
Methods 21, 974–982. resistance in histone methylation-targeted therapy. Nature 627, 221–228.
Wu, S.J., Furlan, S.N., Mihalas, A.B., Kaya-Okur, H.S., Feroze, A.H., Emerson, S.N., Yan, R., Gu, C., You, D., Huang, Z., Qian, J., Yang, Q., Cheng, X., Zhang, L., Wang, H.,
Zheng, Y., Carson, K., Cimino, P.J., Keene, C.D., et al. (2021). Single-cell CUT&Tag Wang, P., et al. (2021). Decoding dynamic epigenetic landscapes in human oocytes
analysis of chromatin modifications in differentiation and tumor progression. Nat using single-cell multi-omics sequencing. Cell Stem Cell 28, 1641–1656.e7.
Biotechnol 39, 819–824. Yang, D., Wang, L.L., Dong, T.T., Shen, Y.H., Guo, X.S., Liu, C.Y., Liu, J., Zhang, P., Li,
Wu, T., Cheng, A.Y., Zhang, Y., Xu, J., Wu, J., Wen, L., Li, X., Liu, B., Dou, X., Wang, J., and Sun, Y.P. (2015). Progranulin promotes colorectal cancer proliferation and
P., et al. (2024b). KARR-seq reveals cellular higher-order RNA structures and angiogenesis through TNFR2/Akt and ERK signaling pathways. Am J Cancer Res
RNA-RNA interactions. Nat Biotechnol 42, 1909–1920. 5, 3085–3097.
Xia, C., Fan, J., Emanuel, G., Hao, J., and Zhuang, X. (2019). Spatial transcriptome Yang, F., Wang, W., Wang, F., Fang, Y., Tang, D., Huang, J., Lu, H., and Yao, J.
profiling by MERFISH reveals subcellular RNA compartmentalization and cell cycle- (2022). scBERT as a large-scale pretrained deep language model for cell type
dependent gene expression. Proc Natl Acad Sci USA 116, 19490–19499. annotation of single-cell RNA-seq data. Nat Mach Intell 4, 852–866.
Xia, K., Sun, H.X., Li, J., Li, J., Zhao, Y., Chen, L., Qin, C., Chen, R., Chen, Z., Liu, G., et Yang, S., Liu, C., Jiang, M., Liu, X., Geng, L., Zhang, Y., Sun, S., Wang, K., Yin, J., Ma,
al. (2022). The single-cell stereo-seq reveals region-specific cell subtypes and S., et al. (2024). A single-nucleus transcriptomic atlas of primate liver aging
transcriptome profiling in Arabidopsis leaves. Dev Cell 57, 1299–1310.e4. uncovers the pro-senescence role of SREBP2 in hepatocytes. Protein Cell 15, 98–
Xiao, Y., Wang, J., Li, J., Zhang, P., Li, J., Zhou, Y., Zhou, Q., Chen, M., Sheng, X., Liu, 120.
Z., et al. (2023). An analytical framework for decoding cell type-specific genetic Yang, W., Wang, P., Luo, M., Cai, Y., Xu, C., Xue, G., Jin, X., Cheng, R., Que, J., Pang,
variation of gene regulation. Nat Commun 14, 3884. F., et al. (2023a). DeepCCI: a deep learning framework for identifying cell-cell
Xing, Q.R., Farran, C.A.E., Zeng, Y.Y., Yi, Y., Warrier, T., Gautam, P., Collins, J.J., Xu, interactions from single-cell RNA sequencing data. Bioinformatics 39, btad596.

1280 SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 [Link]
Yang, X., Liu, G., Feng, G., Bu, D., Wang, P., Jiang, J., Chen, S., Yang, Q., Zhang, Y., Yuan, Z., Zhou, Q., Cai, L., Pan, L., Sun, W., Qumu, S., Yu, S., Feng, J., Zhao, H.,
Man, Z., et al. (2023b). GeneCompass: deciphering universal gene regulatory Zheng, Y., et al. (2021). SEAM is a spatial single nuclear metabolomics method for
mechanisms with knowledge-informed cross-species foundation model. bioRxiv, dissecting tissue microenvironment. Nat Methods 18, 1223–1232.
2023.2009.2026.559542. Zachariadis, V., Cheng, H., Andrews, N., and Enge, M. (2020). A highly scalable
Yang, Y., Li, G., Qian, H., Wilhelmsen, K.C., Shen, Y., and Li, Y. (2021). SMNN: Batch method for joint whole-genome sequencing and gene-expression profiling of single
effect correction for single-cell RNA-seq data via supervised mutual nearest cells. Mol Cell 80, 541–553.e5.
neighbor detection. Brief Bioinf 22, bbaa097. Zeisel, A., Hochgerner, H., Lönnerberg, P., Johnsson, A., Memic, F., van der Zwan, J.,
Yang, Z., Li, C., Fan, Z., Liu, H., Zhang, X., Cai, Z., Xu, L., Luo, J., Huang, Y., He, L., et Häring, M., Braun, E., Borm, L.E., La Manno, G., et al. (2018). Molecular
al. (2017). Single-cell sequencing reveals variants in ARID1A, GPRC5A and MLL2 architecture of the mouse nervous system. Cell 174, 999–1014.e22.
driving self-renewal of human bladder cancer stem cells. Eur Urology 71, 8–12. Zeng, H. (2022). What is a cell type and how to define it? Cell 185, 2739–2755.
Yao, H., Gao, C.C., Zhang, D., Xu, J., Song, G., Fan, X., Liang, D.B., Chen, Y.S., Li, Q., Zeng, H., Huang, J., Ren, J., Wang, C.K., Tang, Z., Zhou, H., Zhou, Y., Shi, H.,
Guo, Y., et al. (2023). scm6A-seq reveals single-cell landscapes of the dynamic m6A Aditham, A., Sui, X., et al. (2023a). Spatially resolved single-cell translatomics at
during oocyte maturation and early embryonic development. Nat Commun 14, molecular resolution. Science 380, eadd3067.
315. Zeng, H., Huang, J., Zhou, H., Meilandt, W.J., Dejanovic, B., Zhou, Y., Bohlen, C.J.,
Yazar, S., Alquicira-Hernandez, J., Wing, K., Senabouth, A., Gordon, M.G., Andersen, Lee, S.H., Ren, J., Liu, A., et al. (2023b). Integrative in situ mapping of single-cell
S., Lu, Q., Rowson, A., Taylor, T.R.P., Clarke, L., et al. (2022). Single-cell eQTL transcriptional states and tissue histopathology in a mouse model of Alzheimer’s
mapping identifies cell type-specific genetic control of autoimmune disease. Science disease. Nat Neurosci 26, 430–446.
376, eabf3041. Zeng, X., Jiang, S., Zhong, Z., Yang, X., Chen, Q., Li, J., Zhu, Z., Song, J., and Yang, C.
Ye, F., Wang, J., Li, J., Mei, Y., and Guo, G. (2024a). Mapping cell atlases at the single- (2024). DIRECT: digital microfluidics for isolation-free shared library construction
cell level. Adv Sci 11, e2305449. of single-cell DNA methylome and transcriptome. Small Methods 8, e2301075.
Ye, F., Zhang, G., E., W., Chen, H., Yu, C., Yang, L., Fu, Y., Li, J., Fu, S., Sun, Z., et al. Zeng, Z., Li, Y., Li, Y., and Luo, Y. (2022). Statistical and machine learning methods
(2022). Construction of the axolotl cell landscape using combinatorial hybridiza- for spatially resolved transcriptomics data analysis. Genome Biol 23, 83.
tion sequencing at single-cell resolution. Nat Commun 13, 4228. Zhai, J., Guo, J., Wan, H., Qi, L., Liu, L., Xiao, Z., Yan, L., Schmitz, D.A., Xu, Y., Yu, D.,
Ye, F., Zhang, S., Fu, Y., Yang, L., Zhang, G., Wu, Y., Pan, J., Chen, H., Wang, X., Ma, et al. (2022). Primate gastrulation and early organogenesis at single-cell
L., et al. (2024b). Fast and flexible profiling of chromatin accessibility and total resolution. Nature 612, 732–738.
RNA expression in single nuclei using Microwell-seq3. Cell Discov 10, 33. Zhang, A.W., O’Flanagan, C., Chavez, E.A., Lim, J.L.P., Ceglia, N., McPherson, A.,
Ye, X., Huang, X., Fu, X., Zhang, X., Lin, R., Zhang, W., Zhang, J., and Lu, Y. (2023). Wiens, M., Walters, P., Chan, T., Hewitson, B., et al. (2019a). Probabilistic cell-type
Myeloid-like tumor hybrid cells in bone marrow promote progression of prostate assignment of single-cell RNA-seq for tumor microenvironment profiling. Nat
cancer bone metastasis. J Hematol Oncol 16, 46. Methods 16, 1007–1015.
Yeung, J., Florescu, M., Zeller, P., de Barbanson, B.A., Wellenstein, M.D., and van Zhang, B., He, P., Lawrence, J.E., Wang, S., Tuck, E., Williams, B.A., Roberts, K.,
Oudenaarden, A. (2023). scChIX-seq infers dynamic relationships between histone Kleshchevnikov, V., Mamanova, L., Bolt, L., et al. (2023a). A human embryonic
modifications in single cells. Nat Biotechnol 41, 813–823. limb cell atlas resolved in space and time. bioRxiv, 2022.2004.2027.489800.
Yin, K., Zhao, M., Lin, L., Chen, Y., Huang, S., Zhu, C., Liang, X., Lin, F., Wei, H., Zeng, Zhang, B., Srivastava, A., Mimitou, E., Stuart, T., Raimondi, I., Hao, Y., Smibert, P.,
H., et al. (2022). Well-paired-seq: a size-exclusion and locally quasi-static and Satija, R. (2022a). Characterizing cellular heterogeneity in chromatin state
hydrodynamic microwell chip for single-cell RNA-Seq. Small Methods 6, with scCUT&Tag-pro. Nat Biotechnol 40, 1220–1230.
e2200341. Zhang, F., Wei, K., Slowikowski, K., Fonseka, C.Y., Rao, D.A., Kelly, S., Goodman, S.
Yin, Y., Jiang, Y., Lam, K.W.G., Berletch, J.B., Disteche, C.M., Noble, W.S., Steemers, F. M., Tabechian, D., Hughes, L.B., Salomon-Escoto, K., et al. (2019b). Defining
J., Camerini-Otero, R.D., Adey, A.C., and Shendure, J. (2019). High-throughput inflammatory cell states in rheumatoid arthritis joint synovial tissues by integrating
single-cell sequencing with linear amplification. Mol Cell 76, 676–690.e10. single-cell transcriptomics and mass cytometry. Nat Immunol 20, 928–942.
Yoshihara, K., Shahmoradgoli, M., Martínez, E., Vegesna, R., Kim, H., Torres-Garcia, Zhang, H., Li, J., Ren, J., Sun, S., Ma, S., Zhang, W., Yu, Y., Cai, Y., Yan, K., Li, W., et
W., Treviño, V., Shen, H., Laird, P.W., Levine, D.A., et al. (2013). Inferring tumour al. (2021a). Single-nucleus transcriptomic landscape of primate hippocampal
purity and stromal and immune cell admixture from expression data. Nat Commun aging. Protein Cell 12, 695–716.
4, 2612. Zhang, K., Hocker, J.D., Miller, M., Hou, X., Chiou, J., Poirion, O.B., Qiu, Y., Li, Y.E.,
Yost, K.E., Satpathy, A.T., Wells, D.K., Qi, Y., Wang, C., Kageyama, R., McNamara, K. Gaulton, K.J., Wang, A., et al. (2021b). A single-cell atlas of chromatin accessibility
L., Granja, J.M., Sarin, K.Y., Brown, R.A., et al. (2019). Clonal replacement of in the human genome. Cell 184, 5985–6001.e19.
tumor-specific T cells following PD-1 blockade. Nat Med 25, 1251–1259. Zhang, L., Li, Z., Skrzypczynska, K.M., Fang, Q., Zhang, W., O’Brien, S.A., He, Y.,
You, Y., Fu, Y., Li, L., Zhang, Z., Jia, S., Lu, S., Ren, W., Liu, Y., Xu, Y., Liu, X., et al. Wang, L., Zhang, Q., Kim, A., et al. (2020a). Single-cell analyses inform
(2024). Systematic comparison of sequencing-based spatial transcriptomic mechanisms of myeloid-targeted therapies in colon cancer. Cell 181, 442–459.e29.
methods. Nat Methods 21, 1743–1754. Zhang, L., Yu, X., Zheng, L., Zhang, Y., Li, Y., Fang, Q., Gao, R., Kang, B., Zhang, Q.,
Yu, H., Ai, N., Peng, P., Ke, Y., Chen, X., Li, Y., Zhao, T., Jiang, S., Liu, J., and Jiang, L. Huang, J.Y., et al. (2018). Lineage tracking reveals dynamic relationships of T cells
(2021a). Comprehensive identification of fetal cis-regulatory elements in the in colorectal cancer. Nature 564, 268–272.
human genome by single-cell multi-omics analysis. bioRxiv, Zhang, L., and Zhang, S. (2018). PBLR: an accurate single cell RNA-seq data
2021.2011.2002.466852. imputation tool considering cell heterogeneity and prior expression level of
Yu, L., Wang, X., Mu, Q., Tam, S.S.T., Loi, D.S.C., Chan, A.K.Y., Poon, W.S., Ng, H.K., dropouts. bioRxiv, 379883.
Chan, D.T.M., Wang, J., et al. (2023a). scONE-seq: a single-cell multi-omics method Zhang, M., Eichhorn, S.W., Zingg, B., Yao, Z., Cotter, K., Zeng, H., Dong, H., and
enables simultaneous dissection of phenotype and genotype heterogeneity from Zhuang, X. (2021c). Spatially resolved cell atlas of the mouse primary motor cortex
frozen tumors. Sci Adv 9, eabp8901. by MERFISH. Nature 598, 137–143.
Yu, M., Abnousi, A., Zhang, Y., Li, G., Lee, L., Chen, Z., Fang, R., Lagler, T.M., Yang, Zhang, M., Yang, H., Wan, L., Wang, Z., Wang, H., Ge, C., Liu, Y., Hao, Y., Zhang, D.,
Y., Wen, J., et al. (2021b). SnapHiC: a computational pipeline to identify chromatin Shi, G., et al. (2020b). Single-cell transcriptomic architecture and intercellular
loops from single-cell Hi-C data. Nat Methods 18, 1056–1059. crosstalk of human intrahepatic cholangiocarcinoma. J Hepatol 73, 1118–1130.
Yu, X., Liu, Z., and Sun, X. (2023b). Single-cell and spatial multi-omics in the plant Zhang, Q., He, Y., Luo, N., Patel, S.J., Han, Y., Gao, R., Modak, M., Carotta, S.,
sciences: technical advances, applications, and perspectives. Plant Commun 4, Haslinger, C., Kind, D., et al. (2019c). Landscape and dynamics of single immune
100508. cells in hepatocellular carcinoma. Cell 179, 829–845.e20.
Yu, Z., and Du, F. (2022). AMC: accurate mutation clustering from single-cell DNA Zhang, S., Fang, W., Zhou, S., Zhu, D., Chen, R., Gao, X., Li, Z., Fu, Y., Zhang, Y.,
sequencing data. Bioinformatics 38, 1732–1734. Yang, F., et al. (2023b). Single cell transcriptomic analyses implicate an
Yu, Z., Du, F., and Song, L. (2022). SCClone: accurate clustering of tumor single-cell immunosuppressive tumor microenvironment in pancreatic cancer liver metas-
DNA sequencing data. Front Genet 13, 823941. tasis. Nat Commun 14, 5123.
Yu, Z., Liu, F., Shi, F., and Du, F. (2023c). rcCAE: a convolutional autoencoder Zhang, T., Zhang, Z., Li, L., Dong, B., Wang, G., and Zhang, D. (2023c). GTAD: A
method for detecting intra-tumor heterogeneity and single-cell copy number graph-based approach for cell spatial composition inference from integrated scRNA-
alterations. Brief Bioinf 24, bbad108. seq and ST-seq data. Brief Bioinf 25, bbad469.
Yuan, H., and Kelley, D.R. (2022). scBasset: Sequence-based modeling of single-cell Zhang, W., Zhang, S., Yan, P., Ren, J., Song, M., Li, J., Lei, J., Pan, H., Wang, S., Ma,
ATAC-seq using convolutional neural networks. Nat Methods 19, 1088–1096. X., et al. (2020c). A single-cell transcriptomic landscape of primate arterial aging.
Yuan, Y., and Bar-Joseph, Z. (2020). GCNG: graph convolutional networks for Nat Commun 11, 2202.
inferring gene interaction from spatial transcriptomics data. Genome Biol 21, 300. Zhang, W., Zhang, X., Liu, Q., Wei, L., Qiao, X., Gao, R., Liu, Z., and Wang, X.

[Link] SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 1281
 (2023d). Deconer: a comprehensive and systematic evaluation toolkit for reference- (2018). Deep learning sequence-based ab initio prediction of variant effects on
based cell type deconvolution algorithms using gene expression data. bioRxiv, expression and disease risk. Nat Genet 50, 1171–1179.
2023.2012.2024.573278. Zhou, T., Zhang, R., Jia, D., Doty, R.T., Munday, A.D., Gao, D., Xin, L., Abkowitz, J.L.,
Zhang, X., Hu, C., Huang, C., Wei, Y., Li, X., Hu, M., Li, H., Wu, J., Czajkowsky, D.M., Duan, Z., and Ma, J. (2023). Concurrent profiling of multiscale 3D genome
Guo, Y., et al. (2022b). Robust acquisition of spatial transcriptional programs in organization and gene expression in single mammalian cells. bioRxiv,
tissues with immunofluorescence-guided laser capture microdissection. Front Cell 2023.07.20.549578.
Dev Biol 10, 853188. Zhou, W., Yan, Y., Guo, Q., Ji, H., Wang, H., Xu, T., Makabel, B., Pilarsky, C., He, G.,
Zhang, X., Li, T., Liu, F., Chen, Y., Yao, J., Li, Z., Huang, Y., and Wang, J. (2019d). Yu, X., et al. (2021). Microfluidics applications for high-throughput single cell
Comparative analysis of droplet-based ultra-high-throughput single-cell RNA-Seq sequencing. J Nanobiotechnol 19, 312.
systems. Mol Cell 73, 130–142.e5. Zhou, Y., Bian, S., Zhou, X., Cui, Y., Wang, W., Wen, L., Guo, L., Fu, W., and Tang, F.
Zhang, Y., Chen, H., Mo, H., Hu, X., Gao, R., Zhao, Y., Liu, B., Niu, L., Sun, X., Yu, X., (2020). Single-cell multiomics sequencing reveals prevalent genomic alterations in
et al. (2021d). Single-cell analyses reveal key immune cell subsets associated with tumor stromal cells of human colorectal cancer. Cancer Cell 38, 818–828.e5.
response to PD-L1 blockade in triple-negative breast cancer. Cancer Cell 39, 1578– Zhu, B., Chen, S., Bai, Y., Chen, H., Liao, G., Mukherjee, N., Vazquez, G., McIlwain, D.
1593.e8. R., Tzankov, A., Lee, I.T., et al. (2023). Robust single-cell matching and multimodal
Zhao, S., Zhang, J., and Nie, Z. (2023). Large-scale cell representation learning via analysis using shared and distinct features. Nat Methods 20, 304–315.
divide-and-conquer contrastive learning. arXiv, 2306.04371. Zhu, C., Yu, M., Huang, H., Juric, I., Abnousi, A., Hu, R., Lucero, J., Behrens, M.M.,
Zhao, T., Chiang, Z.D., Morriss, J.W., LaFave, L.M., Murray, E.M., Del Priore, I., Meli, Hu, M., and Ren, B. (2019). An ultra high-throughput method for single-cell joint
K., Lareau, C.A., Nadaf, N.M., Li, J., et al. (2022). Spatial genomics enables multi- analysis of open chromatin and transcriptome. Nat Struct Mol Biol 26, 1063–
modal study of clonal heterogeneity in tissues. Nature 601, 85–91. 1070.
Zheng, C., Fass, J.N., Shih, Y.P., Gunderson, A.J., Sanjuan Silva, N., Huang, H., Zhu, C., Zhang, Y., Li, Y.E., Lucero, J., Behrens, M.M., and Ren, B. (2021). Joint
Bernard, B.M., Rajamanickam, V., Slagel, J., Bifulco, C.B., et al. (2022). profiling of histone modifications and transcriptome in single cells from mouse
Transcriptomic profiles of neoantigen-reactive T cells in human gastrointestinal brain. Nat Methods 18, 283–292.
cancers. Cancer Cell 40, 410–423.e7. Zhu, H., Galdos, F.X., Lee, D., Waliany, S., Huang, Y.V., Ryan, J., Dang, K., Neal, J.W.,
Zheng, C., Zheng, L., Yoo, J.K., Guo, H., Zhang, Y., Guo, X., Kang, B., Hu, R., Huang, J. Wakelee, H.A., Reddy, S.A., et al. (2022). Identification of pathogenic immune cell
Y., Zhang, Q., et al. (2017a). Landscape of infiltrating T cells in liver cancer subsets associated with checkpoint inhibitor-induced myocarditis. Circulation 146,
revealed by single-cell sequencing. Cell 169, 1342–1356.e16. 316–335.
Zheng, G.X.Y., Terry, J.M., Belgrader, P., Ryvkin, P., Bent, Z.W., Wilson, R., Ziraldo, S. Zhu, J., and Sabatti, C. (2020). Integrative spatial single-cell analysis with graph-
B., Wheeler, T.D., McDermott, G.P., Zhu, J., et al. (2017b). Massively parallel digital based feature learning. bioRxiv, 2020.2008.2012.248971.
transcriptional profiling of single cells. Nat Commun 8, 14049. Zhu, Y.Y., Machleder, E.M., Chenchik, A., Li, R., and Siebert, P.D. (2001). Reverse
Zheng, L., Qin, S., Si, W., Wang, A., Xing, B., Gao, R., Ren, X., Wang, L., Wu, X., transcriptase template switching: a SMARTTM approach for full-length cDNA
Zhang, J., et al. (2021). Pan-cancer single-cell landscape of tumor-infiltrating T library construction. BioTechniques 30, 892–897.
cells. Science 374, abe6474. Zong, C., Lu, S., Chapman, A.R., and Xie, X.S. (2012). Genome-wide detection of
Zheng, W., Min, W., Wang, S., and Boeva, V. (2023). TsImpute: An accurate two-step single-nucleotide and copy-number variations of a single human cell. Science 338,
imputation method for single-cell RNA-seq data. Bioinformatics 39, btad731. 1622–1626.
Zhong, Y., Wan, Y.W., Pang, K., Chow, L.M., and Liu, Z. (2013). Digital sorting of Zou, B., Zhang, T., Zhou, R., Jiang, X., Yang, H., Jin, X., and Bai, Y. (2021). deepMNN:
complex tissues for cell type-specific gene expression profiles. BMC Bioinf 14, 89. deep learning-based single-cell RNA sequencing data batch correction using
Zhou, F., Wang, R., Yuan, P., Ren, Y., Mao, Y., Li, R., Lian, Y., Li, J., Wen, L., Yan, L., mutual nearest neighbors. Front Genet 12, 708981.
et al. (2019). Reconstituting the transcriptome and DNA methylome landscapes of Zou, Z., Zhang, C., Wang, Q., Hou, Z., Xiong, Z., Kong, F., Wang, Q., Song, J., Liu, B.,
human implantation. Nature 572, 660–664. Liu, B., et al. (2022). Translatome and transcriptome co-profiling reveals a role of
Zhou, J., Theesfeld, C.L., Yao, K., Chen, K.M., Wong, A.K., and Troyanskaya, O.G. TPRXs in human zygotic genome activation. Science 378, abo7923.

1282 SCIENCE CHINA Life Sciences Vol.68 No.5, 1226–1282 May 2025 [Link]