How SynPlexity uses BMS to map sequence variation and drive biological discovery

We are excited to take part in our annual Symposium! What topics will we be discussing? Collaboratorium Annual Symposium: Causality in Biology & AI 💻 When analyzing biological data with AI, simply identifying associations isn’t enough. Correlations can be misleading because sometimes two events occur together without one actually causing the other. => Understanding the true causal relationships helps us build models that not only fit existing data but can also reliably predict outcomes in new situations or with new datasets. 🔗 Having a clear grasp of causal links also makes AI models more interpretable and trustworthy. Instead of being a black box that produces results without explanation, the model becomes something scientists can validate and have confidence in. Moreover, it helps uncover hidden biases or confounding factors in the data, ensuring that the conclusions we draw are solid and reliable. 🔬 In biology, we often work with diverse types of data — from genes to proteins to entire organisms. Understanding how these components interact causally allows us to integrate this information in a way that reflects real biological processes, rather than just coincidental patterns. Ultimately, focusing on causality is what makes models truly useful and robust, enabling us to design better experiments, make accurate predictions, and drive scientific discovery forward. 👏 We will explore all of this during our symposium on November 10th and 11th! #BCNCollaboratorium #CRG #EMBLBCN #SYMPOSIUM #PRBB

Today, I am excited to announce the launch of BioModelBench.com, the "HuggingFace" of Bio-models, to help scientists more easily discover, compare, and deploy computational biology tools, datasets, and workflows. BioModelBench's database is the product of nearly 4 years of monitoring and mapping the computational biology and bio+AI landscape. Its AI search feature allows you to input information related to your experiments and find the most relevant and cutting edge tools and datasets. Our AI Assistants and AI Scientist take this further to help you understand best uses for specific resources and to visualize+automate your stack. Initially, AI feature use will be free, but with credits limited on a daily first-come-first-serve basis, so get there early! If you miss out, or if you're just not sure where to start, drop us a message, and we'll work with you to get familiar with BioModelBench and support your computational biology deployments! Biomodelbench.com by Ourobio #computationalbiology #AI #AIscientist #synbio #syntheticbiology #HF #Bioinformatics #OpenScience

Big move from Anthropic. Claude has officially entered the life sciences world with the launch of Claude for Life Sciences. It builds on Anthropic’s existing models and now connects directly with scientific tools like Benchling, PubMed, Wiley’s Scholar Gateway, Synapse.org, and 10x Genomics. Researchers can use it to run literature reviews, generate hypotheses, draft study protocols, and even process genomic data through Claude Code. It can also help with regulatory submissions and compliance documentation. James Zou from Stanford called Claude “extremely valuable” in developing Paper2Agent, a project turning static research papers into interactive AI agents that act as co-scientists. From a market perspective, this is a clear sign that frontier LLMs are starting to go vertical. Instead of broad, general-purpose tools, we’re seeing deep integrations with sector-specific workflows. For AI professionals, it shows where the next wave of demand will come from specialists who can translate foundation model capability into real scientific and commercial outcomes. Would be great to hear your thoughts ⬇️ DeepRec.ai Trinnovo Group | B Corp™ #llm #claude #anthropic #lifesciences #news

Very excited to see the implications this has for 10x Genomics and Opentrons Labworks Inc. The technical barrier to entry for cutting-edge research tools is lowering FAST! Millennium Science 💛 10x Genomics

This is a major step forward for AI in life sciences. Having worked in biotech before moving into design and creative tech, I find this collaboration between Anthropic and 10x Genomics especially exciting. It shows how AI can make complex science more intuitive — turning data, papers, and workflows into accessible, human-centered tools. What stands out most is the shift from broad AI to deep, domain-specific integrations. That’s where real innovation — and impact — happens. Excited to see how these kinds of partnerships will reshape research, collaboration, and discovery across disciplines. #AI #LifeSciences #Biotech #Innovation #DesignThinking #Claude #Benchling

gReLU: Unifying deep learning for DNA sequence modeling and design Deep learning has transformed genomics — from predicting regulatory activity to designing synthetic DNA. But progress has been fragmented: every new model comes with custom code, incompatible data pipelines, and isolated tools for interpretation or design. Avantika Lal and coauthors introduce gReLU, a comprehensive open-source framework that brings all of this under one roof. Built in PyTorch, gReLU integrates data preprocessing, model training, interpretation, variant effect prediction, and even regulatory element design. It supports convolutional and transformer architectures alike, includes a model zoo, and provides built-in tools for sequence attribution, motif discovery, and directed evolution of DNA. Researchers can now train, interpret, and design sequences in a unified workflow — reducing friction and boosting reproducibility across labs. The framework goes beyond prediction: it can visualize long-range enhancer–gene interactions, test variant effects in silico, and guide the optimization of regulatory DNA for specific cell types. This is a big step toward “foundation models” for genomics — modular, interoperable, and open tools that make biological sequence modeling as seamless as training large language models. Paper: https://lnkd.in/du5QRYpn #AIforScience #Genomics #DeepLearning #Bioinformatics #ComputationalBiology #SyntheticBiology #DNA #MachineLearning #gReLU #Genentech #AIinBiology #Transformers #OpenSource #RegulatoryGenomics

The article “We Still Can’t Predict Much of Anything in Biology” by Claus Wilke, circulated among our team this week, validating what our experts in synthetic biology already know: data is the key to moving biology forward. We know this as a fact. The idea that we can simply model biological systems today ignores reality. We’re still a long way from tuning a few parameters and producing the biology we want. At the Oregon Biosciences Annual Conference, I had a chance to listen in on a roundtable discussion on AI in health and biology. The most important takeaway was clear: data must expand dramatically, and its structure matters as much as its volume. The form of the tensor defines what the model can learn, and the inclusion of negative data, where “no” is as valuable as “yes”, is essential for progress in biological prediction. This is where SynPlexity plays a critical role. Our platform leverages Broad Mutational Scanning (BMS) to generate the high-dimensional, information-rich datasets that discovery biology needs. We expect more advanced analytical tools will evolve to fully capitalize on BMS, but the fundamental truth remains: biology’s complexity demands more and better data.

Just finished reading about BoltzGen, and it got me thinking — If a model can generalize across protein folds and create nanomolar binders with a 67% success rate… what does “innovation” even mean in molecular biology anymore? We’ve always treated structure prediction as the end goal. But what happens when the model starts to design? When geometry-based representations outperform traditional biochemical reasoning — are we still “understanding” biology, or are we just guiding a generative process we can’t fully interpret? Another question: How far are we from a closed-loop system — AI → wet lab validation → retraining → autonomous discovery? And if that loop runs faster than human hypothesis cycles, what’s the role of a scientist then — observer or optimizer? Curious what you all think. Are we entering the age of synthetic intuition in biology? #AI #ProteinDesign #BoltzGen #MolecularML #SyntheticBiology

Literature Review AI: The hardest part of any research project? The beginning. That endless cycle of trial-and-error can last for months, even years. For me, it was 1.5 years of struggling with my bioinformatics research on hypothetical proteins. That's why I created the Literature Review AI—a tool designed to turn that initial confusion into a clear, strategic map in minutes. In my latest blog post, I share the personal story behind this tool and how it can help researchers move from a broad question to a focused hypothesis, fast. Stop drowning in data and start discovering. Read the full story and see how it works. The article is now live on my website: https://lnkd.in/gFZfguzk #Research #Innovation #AI #Bioinformatics #ProblemSolving #Tech #Academia

Large Language Models (LLMs) are transforming how we write, communicate, and code. More recently, they have also begun to impact how scientific research is conducted, but LLMs alone are not enough. At Kitware Inc., we built Viime-Extract to drive new scientific discoveries in infectious diseases. Viime-Extract is an LLM-powered system that turns vast biomedical literature into structured, rich, and easier-to-digest knowledge graphs. Using an LLM is easy, but obtaining meaningful results for a specific scientific domain requires a deep understanding of the application context. In collaboration with world-class metabolomics experts, we developed an engineering layer that combines intelligent text preprocessing, specialized prompt design, fuzzy matching and entity disambiguation, as well as integration with biomedical ontologies. Most importantly, by keeping a human in the loop, we ensure that Viime-Extract’s outputs are meaningful, accurate, and useful for researchers. Our collaboration between engineers and subject matter experts ensures that these tools do not just summarize papers but generate real insight, helping researchers navigate millions of publications and uncover new biological connections. Read the full story on our blog: https://lnkd.in/e9vWEV2v cc. Roni Choudhury Thomas O'Connell, Ph.D. #AI #BiomedicalResearch #KnowledgeGraphs #LLMs #Bioinformatics #DataScience #DataVisualization #OntologyDatabases #NIAIDFunded #PromptEngineering #Metabolomics #ResearchInnovation