Trends in Food Science & Technology 20 (2009) 557e566

Review

Metabolomic analysis
in food science: Metabolomic analyses have been generally classified as
targeted or untargeted (Fig. 1). Targeted analyses focus on
a specific group of intended metabolites with most cases
a review requiring identification and quantification of as many me-
tabolites within the group (Ramautar, Demirci, & Jong,
2006). Targeted analyses are important for assessing the
Juan M. Cevallos-Cevallosa, behavior of a specific group of compounds in the sample
José I. Reyes-De-Corcueraa,*, under determined conditions. Targeted metabolomics typi-
cally requires higher level of purification and a selective
Edgardo Etxeberriaa, extraction of metabolites. In contrast, untargeted (a.k.a.
comprehensive) metabolomics focuses on the detection of
Michelle D. Danyluka and as many groups of metabolites as possible to obtain patterns
Gary E. Rodrickb or fingerprints without necessarily identifying nor quantify-
ing a specific compound(s) (Monton & Soga, 2007). Untar-
a
University of Florida, IFAS, Citrus Research and geted analyses have been used in the identification of
Education Center, 700 Experiment Station Rd., Lake possible fingerprints of biological phenomena such as plant
Alfred, FL 33850, USA (Tel.: D1 863 956 1151; diseases (Cevallos-Cevallos, Rouseff, & Reyes-De-Cor-
fax: D1 863 956 4631; e-mail: [email protected]) cuera, 2009). Based on the specific objective of the analysis
b and data manipulation, most metabolomic studies can also
University of Florida, Department of Food Science
be classified as discriminative, informative, and/or predic-
and Human Nutrition, 359 Food Science and Human
tive (Fig. 1). Discriminative analyses have been aimed to
Nutrition Building, P.O. Box 110370, Newell Drive,
find differences between sample populations without neces-
Gainesville, FL 32611, USA
sarily creating statistical models or evaluating possible
pathways that may elucidate such differences. Wine has
Metabolomics has emerged as an important tool in many been classified by grape variety and production area by me-
disciplines such as human diseases and nutrition, drug discovery, tabolomic techniques (Son et al., 2008). Discrimination is
plant physiology and others. In food science, metabolomics usually achieved by the use of multivariate data analysis
has recently risen as a tool for quality, processing and safety (MVDA) techniques intended to maximize classification,
of raw materials and final products. This article discusses the principal components analysis (PCA) being the most used
latest advances in food metabolomics from the discriminative, tool. PCA and other MVDA tools have been widely
predictive, and informative approaches, as well as the typical described in other reviews (Kemsley et al., 2007; van der
methods used at each step of the metabolomic analysis. Werf, Jellema, & Hankemeier, 2005). In contrast, informa-
tive metabolomic analyses have focused on the identifica-
tion and quantification of targeted or untargeted
metabolites to obtain sample intrinsic information. Infor-
Introduction
mative metabolomics has been used in the development
Metabolomics, the study of ‘‘as-many-small-metabo-
and continuous update of metabolite databases such as
lites-as-possible’’ in a system, has become an important
the human metabolome database (Wishart et al., 2007).
tool in many research areas. Recent reviews and perspec-
Possible pathways, discovery of novel bioactive com-
tives in the areas of human diseases (Kaddurah-Daouk &
pounds, discovery of biomarkers, creation of specialized
Krishnan, 2009), drug discovery (Wishart, 2008a), plant
metabolite databases, and metabolites functionality studies
analysis (Hall, Brouwer, & Fitzgerald, 2008), human nutri-
can also be carried out by informative metabolomics.
tion (Wishart, 2008b), and others, have shown the broad
Finally, some metabolomics reports have been predictive.
impact and rapid growth of metabolomics.
In this case, statistical models based on metabolite profile
and abundance are created to predict a variable that is dif-
* Corresponding author. ficult to quantify by other means. Metabolite-based models
0924-2244/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.tifs.2009.07.002
558 J.M. Cevallos-Cevallos et al. / Trends in Food Science & Technology 20 (2009) 557e566

METABOLOMICS SAMPLE

Extraction &
sample preparation Preparation
Grinding, freeze-drying,
dilution, etc.

Targeted Untargeted
Extraction
Targeted or untargeted
Data treatment
Derivatization Separation
mainly for GC analysis LC, GC, CE

Discriminative Predictive Informative

Data treatment
for prediction of green tea sensory quality have been devel- Peak identification, Alignment, Statistical
analysis: ANOVA, PCA, PLS, LDA, CN, RF, HCA
oped (Ikeda, Kanaya, Yonetani, Kobayashi, & Fukusaki,
2007). These models are usually produced by partial least Fig. 2. Schematic representation of the process of metabolomic
square (PLS) regression as discussed in Data treatment sec- analysis.
tion of this review.
In food science, metabolomics has the potential for solv- Proper grinding enhances the release of metabolites during
ing major problems worldwide as it is being applied in food extraction. Freeze-drying acts as a concentration step and
research programs such as the Metabolomics for Plants, minimizes possible differences in metabolites due to dis-
Health and OutReach (METHA-PHOR) initiative (Hall, similarities in moisture content between groups of sample.
2007). Moreover, metabolomics is considered an efficient Other concentrated liquid samples such as honey can be
tool for addressing future needs in agriculture (Green, Qur- diluted as a preliminary step (Donarski, Jones, & Charlton,
eshi, Long, Burfening, & Hamernik, 2007) and human nu- 2008). However, to maximize the amount of information to
trition (Green et al., 2007; Hall et al., 2008). be collected, concentration steps are more suitable. For ex-
Discriminative, informative, and predictive metabolo- ample, metabolites in wine (Son et al., 2008) and volatiles
mics have been recently used in combination for quality, in olive oil (Cavaliere et al., 2007) have been concentrated
nutrition, and food components analysis (Wishart, 2008b) by lyophilization and solid phase microextraction (SPME)
with a significant expansion to other food applications in respectively.
the last two years. This paper presents an in depth review
of recent metabolomics studies in food from the perspective Extraction
of the extraction, separation, detection, and data treatment, The initial extraction procedure is aimed at maximizing
as well as the application of discriminative, informative, the amount and concentration of the compounds of interest.
and predictive metabolomics in the areas of food quality, For this reason, extraction is probably the most critical step
safety, regulations, microbiology, and processing. in metabolomics. In untargeted metabolomics, the nature of
compounds of interest is mostly unknown. Hence, several
solvents and extraction methods should be tested and com-
The process of metabolomic analysis
pared between the groups of samples. Most reports on un-
Metabolomic analyses consist of a sequence of steps in-
targeted food analysis do not describe preliminary
cluding sample preparation, metabolite extraction, derivati-
comparisons among extraction solvents tested. However,
zation, metabolite separation, detection, and data treatment
the extraction methods used in foods have been similar to
(Fig. 2). However, not every steps is always needed. Only
those found optimal in comparable research fields such as
detection and data analysis have been essential steps in
non-food plant metabolomics. For instance, the combina-
all reported metabolomics studies. The selection of the
tion methanolewaterechloroform (MeOHeH2OeCHCl3)
steps depends on the type of study (untargeted vs. targeted),
in different proportions was shown to be superior to other
kind of sample (e.g. solids vs. liquids), instrumentation to
solvents for untargeted studies in plants such as Arabidop-
be used for separation (e.g. GC vs. LC) and detection
sis thaliana (Gullberg, Jonsson, Nordstrom, Sjostrom, &
method (e.g. MS vs. NMR). Table 1 summarizes recent me-
Moritz, 2004) because of its capacity of extracting both
tabolomics studies used for food analysis.
hydrophilic and hydrophobic compounds. Therefore, the ef-
fectiveness of MeOHeH2OeCHCl3 in green tea (Pongsuwan
Sample preparation et al., 2008), potatoes (Dobson et al., 2008) and other foods
Solid samples such as apple peel (Rudell, Mattheis, & was anticipated. For untargeted analysis, the use of sequen-
Curry, 2008) and potatoes (Dobson et al., 2008) are typi- tial and selective extractions followed by metabolite analy-
cally ground under liquid nitrogen or after freeze-drying. sis of each extract was previously recommended (Dixon
Table 1. Most common metabolomics processes in food analysis.

Sample: Purpose of analysis Type Extraction and preparation Separationedetection Data treatment Reference
Apples: light induced changes in peel Untargeted/discriminative MeOH GCeMS PCA Rudell et al., 2008
Derivatization for GCeMS LCeMS
Berries: polyphenol composition Targeted/informative Acetic acid þ water LCeMS Compound McDougall et al., 2008
C18 and Sephadex LH 20 columns DIMS identification
Broccoli, mustard, and brassica: Targeted/informative Hot water (90  C) þ sonication LCeMSn Compound Rochfort et al., 2008
glucosinolates composition identification

J.M. Cevallos-Cevallos et al. / Trends in Food Science & Technology 20 (2009) 557e566
Broccoli: variety differentiation Untargeted/discriminative Freeze dried MeOH þ H2O LCeUVeMS PCA, ANOVA Luthria et al., 2008
DIMS
Cheese: Production control Untargeted/informative e IMS Compound Vautz et al., 2006
identification
E. coli: glycolisis metabolites Targeted/informative Indirect thermal treatment LCeMS Compound Schaub & Reuss, 2008
identification
Ginseng: variety differentiation Untargeted/discriminative Deuterated MeOH þ buffered water NMR PCA Kang et al., 2008
Green: tea quality Untargeted/predictive Freeze dried MeOH þ H2O þ CHCl3 UPLCeTOFeMS PCA, PLS Pongsuwan et al., 2008
Honey: origin verification Untargeted/discriminative/ Buffered water NMR PLSeGP Donarski et al., 2008
predictive
Maize: GMO identification Untargeted/discriminative MeOH þ water þ ultrasonication CEeTOFeMS Student’s t, PCA Levandi et al., 2008
Meat: quality/safety Untargeted/discriminative Neutral desorption EESIeMS PCA Chen et al., 2007
Olive oil: origin differentiation Targeted/discriminative SPME GCeCIeMS LDA KruskaleWallis Cavaliere et al., 2007
and WaldeWolfowitz
tests
Pine mushrooms: quality Untargeted/discriminative MeOH þ H2O þ CHCl3 NMR PCA Cho et al., 2007
differentiation
Potato: GM differentiation Untargeted/discriminative MeOH þ H2O þ CHCl3 GCeMS PCA Catchpole et al., 2005
Derivatization for GCeMS DIMS
Potato: identification of cultivars Untargeted/discriminative/ Freeze dried þ MeOH þ water þ GCeTOFeMS ANOVA, PCA Dobson et al., 2008
informative chloroform þ derivatization
Potato: variety differentiation Untargeted/discriminative/ MeOH þ H2O þ CHCl3 GCeMS RF Beckmann et al., 2007
informative Derivatization for GCeMS DIMS
Soybean: GMO differentiation Untargeted/informative MeOHeEtOHeH2O CEeTOFeMS Compound Garcia-Villalba et al.,
identification 2008
Spinach: E. coli contamination Untargeted/discriminative Neutral desorption EESIeMS PCA Chen et al., 2007
Tomato paste: changes during Targeted to antioxidants/ Targeted: H2OeMeOH and MeOHeCHCl3 LCeantioxidant ANOVA, PCA Capanoglu et al., 2008
production informative detector
Untargeted/informative Untargeted: Formic acideMeOHeH2O LCeTOFeMS
Tomato: metabolite correlations Untargeted/predictive Volatiles: EDTAeNaOHeH2O þ SPME GCeMS PCA, LDA, CN Ursem et al., 2008
Sugars and organic acids: MeOH þ derivatization
Tomato: variety differentiation Untargeted/discriminative Lyophilization þ MeOH þ sonication LCeTOFeMS PCA Moco et al., 2008
NMR
Tomato: volatiles analysis Targeted/discriminative EDTAeNaOHeH2O þ SPME GCeMS PCA, HCA Tikunov et al., 2005
Watermelon: quality evaluation Untargeted/predictive Buffered D2O NMR PLSeLDA Tarachiwin et al., 2008
Wine: metabolite characterization Untargeted/discriminative Lyophilized þ buffered D2O NMR PCA, PLS Son et al., 2008
Yeast: aroma compounds production Targeted/discriminative Diethyl ether GCeFID PCA, PLS Rossouw et al., 2008
Yeast: strain differentiation Untargeted/discriminative Lyophilization þ derivatization GCeTOFeMS PCA, HCA MacKenzie et al., 2008
Yeast: strain differentiation Untargeted/discriminative e NIR PCA Cozzolino et al., 2006
LDA

559
560 J.M. Cevallos-Cevallos et al. / Trends in Food Science & Technology 20 (2009) 557e566

et al., 2006). Usually, an initial hydrophilic extraction (typ- detection of compounds of interest. In food metabolomic
ically with MeOHeH2O) followed by centrifugation and analysis, several silylation reactions have been carried out
hydrophobic extraction (typically with CHCl3) of the pellet at 37  C for 90 min (Beckmann et al., 2007; Dobson et al.,
achieves this purpose. Sequential extraction maximized the 2008) with good results.
amount of metabolites from tomato paste (Capanoglu,
Beekwilder, Boyacioglu, Hall, & De Vos, 2008) finding dis-
criminating compounds in both hydrophilic and hydropho- Separation and detection
bic fractions. Conversely, analysis of other food stuff such Separation and detection of the metabolites have been
as potato (Dobson et al., 2008) and mushrooms (Cho, Kim, considered the key steps in metabolic profiling. Particular
& Choi, 2007) showed low or no sample discrimination in attention has been given to separation techniques such as
the hydrophobic fractions. Similar observations made in liquid chromatography (LC) in its high performance
other areas such as analysis in plant leaves (Cevallos-Ceval- (HPLC) or ultra performance (UPLC) forms, gas chroma-
los et al., 2009) suggest a higher suitability of hydrophilic tography (GC), capillary electrophoresis (CE), as well as
extracts for discriminative metabolomic analyses. Hydro- the coupling of these instruments to detection techniques
philic extractions in untargeted food analysis such as apple such as mass spectrometry (MS), nuclear magnetic reso-
(Rudell et al., 2008) and broccoli (Luthria et al., 2008) have nance (NMR), and near infrared spectrometry (NIR). Work-
usually been carried out by MeOH or MeOHeH2O. Other ing principle as well as individual and hyphenated
extractions based on deuterium oxide (D2O) for NMR anal- suitability of these techniques in metabolomics have been
ysis are also common. Novel methods for extraction of me- broadly discussed (Bedair & Sumner, 2008; Rochfort,
tabolites from frozen meat, where a desorption gas hits the 2005; Toyo’oka, 2008; Wishart, 2008b).
meat surface extracting metabolites further carried to the In food metabolomics most separation analyses have
ionization and detection chambers, have been reported been achieved by GC, CE, and LC as seen in Table 1. Com-
(Chen, Wortmann, & Zenobi, 2007). Extraction for targeted parison and suitability of these techniques in food analysis
analysis relies on previous knowledge of the analytes na- have been discussed in other reviews (Wishart, 2008b).
ture. Polyphenols have been extracted from berries with Among non-conventional techniques, ion mobility spec-
a watereacetic acid combination (McDougall, Martinus- trometry (IMS), where food metabolites are carried in
sen, & Stewart, 2008), and hot water was used for targeted a flow of inert gas, ionized, and separated by a drift gas
analysis of glucosinolates in broccoli and mustard seeds flowing in the opposite direction, has been applied to me-
(Rochfort, Trenerry, Imsic, Panozzo, & Jones, 2008). To tabolomic analysis of cheese, beer, and food packaging ma-
maximize the number and amount of metabolites obtained terial (Vautz et al., 2006). Detection methods are mostly
and to reduce extraction time, disruption methods such as based on UV, NIR, MS, or NMR techniques. In food metab-
ultrasonic treatments are usually part of both untargeted olomics MS and NMR have been used the most (Table 1).
and targeted extractions. A greater amount of data is generally obtained by using MS
accompanied by high throughput separation techniques
Derivatization such as HPLC or UPLC as shown in Table 2. For instance,
In food metabolomics, derivatization is commonly used the quality of green tea has been evaluated by NMR (Tara-
prior to GC analysis in order to increase volatility of analytes. chiwin, Ute, Kobayashi, & Fukusakii, 2007) and UPLCe
Derivatization is usually a two-step process starting with MS (Pongsuwan et al., 2008) and statistical models from
oximation (conversion of aldehydes and ketones to oximes) UPLCeMS yielded a higher prediction coefficient than
of the sample to reduce tautomerism (especially from mono- models from NMR, probably due to the higher number of
saccharide), followed by silylation to increase volatility by peaks detected by UPLCeMS. However, other factors
reducing hydrophilicity of functional groups OH, SH or
NH (Gullberg et al., 2004). Several oximation and silylation Table 2. Examples representative of common number of peaks
reagents have been tested in the past. Gullberg et al. (2004) reported in food metabolomics.a
reviewed previous comparisons among derivatization re-
Technique Peaks reported Main references
agents and reported that methoxiamine hydrochloride in pyr-
idine and N-methyl-N-trimethylsilyltrifluoroacetamide were HPLCeUV 40 detected Defernez et al., 2004
UPLCeMS 1560 detected Pongsuwan et al., 2008
the most appropriate reagents for oximation and silylation re- GCeMS 91e142 detected Beckmann et al., 2007;
spectively. In food analyses, these reagents have shown to MacKenzie et al., 2008
improve GC separation of metabolites from potato (Beck- CEeMS 27e45 detected Garcia-Villalba et al.,
mann, Enot, Overy, & Draper, 2007) and other products. 2008; Levandi et al., 2008
Derivatization times and temperatures affect each metabolite NMR 16e20 identified Jahangir, Kim, Choi, &
Verpoorte, 2008;
independently with major changes at the beginning of the Son et al., 2008
reaction (Ma, Wang, Lu, Xu, & Liu, 2008). Therefore, pre- a
liminary experiments should be done to determine optimum Food matrix and extraction methods greatly influence the num-
ber of detected peaks.
derivatization times and temperatures that maximize the
 J.M. Cevallos-Cevallos et al. / Trends in Food Science & Technology 20 (2009) 557e566 561

such as sample variability should also be considered. Al- comparison without dimensionality reduction. RF has
though not as sensitive as the other detection techniques, allowed classification of potato varieties by pairwise com-
NIR has provided a fast non-destructive fingerprint in sev- parisons with accuracy values greater than 92%. Also cre-
eral metabolomic analyses such as strain differentiation of ation of a Mastermix potato model allowed discrimination
wine yeast (Cozzolino, Flood, Bellon, Gishen, & Lopes, of a larger number of potato varieties through RF (Beck-
2006). Another technique, direct infusion mass spectrome- mann et al., 2007). Table 1 shows that most of the
try (DIMS) methods do not require a previous separation MVDA in food have relied on PCA, PLS, and other linear
step achieving faster results as applied for broccoli (Luthria methods. Non-linear structures associated with the data
et al., 2008). were not considered. Non-linear methods for dimensional-
ity reduction have been shown to outperform linear MVDA
Data treatment tools in areas such as gene and protein expression studies
Metabolomic data have usually been submitted to com- (Lee, Rodriguez, & Madabhushi, 2008). However, to the
pound identification and MVDA. Compound identification best of our knowledge, application of non-linear methods
has been mainly achieved by database matching and com- for metabolomics data analysis has not been reported in
parison with pure standards ran under same conditions. foods. Non-linear PCA, Self Organizing Map (SOM), visu-
Data analysis in food metabolomics is largely carried out alization induced SOM, multidimensional scaling, and
by several chemometrics tools. Typically, metabolomics other non-linear tools that have the potential to be applied
data have been aligned before comparison to correct for to foods have been recently reviewed (Yin, 2007).
instrumental deviations on retention/migration times.
Alignment has been shown to drastically improve the Metabolomics in food quality
performance of MVDA techniques (Son et al., 2008). Targeted metabolomics focused on volatiles has shown
Examples of alignment software include MetAlign for great potential to assess pre-harvest issues that may affect
LCeMS and GCeMS (Sumner, Urbanczyk-Wochniak, & quality. Pre-harvest fungal diseases in mango (Moalemiyan,
Broeckling, 2008) or alignment tools of 32 Karat for CE. Vikram, & Kushalappa, 2007), post-harvest bacterial con-
Alignment routines written in MATLAB (The MathWorks tamination of onions (Vikram, Hamzehzarghani, & Kusha-
Natick, MA) have been reported. Discriminative metabolo- lappa, 2005) and McIntosh apples (Vikram, Prithiviraj,
mics usually relies on multivariate methods such as PCA Hamzehzarghani, & Kushalappa, 2004), as well as diseases
for sample grouping. PCA creates new variables (principal of stored carrots (Vikram, Lui, Hossain, & Kushalappa,
components) by linear combinations of the metabolites de- 2006) have been assessed by sampling headspace metabo-
tected while maximizing sample variation. Grouping occurs lites followed by GCeMS analysis. In each case, the vola-
when comparing the values of two or more principal com- tile profile was found to be disease-specific, and several
ponents of each sample as used for discrimination of broc- compounds were tentatively identified by GCeMS data-
coli varieties (Luthria et al., 2008). On the contrary, PLS is bases. Additionally, changes in polyphenolic compounds
a MVDA technique that allows sample discrimination by during berries breeding (Stewart et al., 2007) have been
reduction of dimensionality while maximizing correlation characterized by informative metabolomics. Post-harvest
between variables. PLS has been the main technique used metabolomic analysis has the potential for detection and
for predictive metabolomics studies such as the creation understanding food spoilage as reviewed by Kushalappa,
of a metabolite-based model for sensory evaluation of wa- Vikram, and Raghavan (2008).
termelon (Tarachiwin, Masako, & Fukusaki, 2008). Simi- The development of novel metabolomic techniques such
larly, linear discriminant analysis (LDA) with a priori as IMS has permitted monitoring of quality attributes dur-
classification hypothesis was used for discrimination of ol- ing food processing. Because IMS allows in situ automatic
ive oil according to origin (Cavaliere et al., 2007). Reviews sampling, it can be used for determining the completion of
on PCA, PLS, and LD are widely found in the literature certain processes assuring standard quality based in a group
(Kemsley et al., 2007; van der Werf et al., 2005). Also, cor- of metabolites. This type of analysis fits the needs of bio-
relation techniques such as correlation network (CN) anal- technological food processes in which metabolites are
ysis have successfully lead to determining the link between changing with time. Targeted informative (concentration
metabolites and establish possible reactions in several in- aimed) IMS has been applied for the detection of diacetyl
formative metabolomics studies such as genotype differen- and 2,3 pentadione compounds in beer to determine the
tiation of tomatoes (Ursem, Tikunov, Bovy, van Berloo, & endpoint of the fermentation (Vautz et al., 2006). Discrim-
van Eeuwijk, 2008). Genetic programming (GP) is another inative metabolomics has allowed classification of health
discriminative tool that was utilized to improve the sensitiv- supplements based on their quality and origin (Kooy,
ity and selectivity of the PLS models for honey origin de- Verpoorte, & Meyer, 2008; Liu, Si, Wan, Lin, & Xu, 2008).
termination (Donarski et al., 2008). Most of the MVDA Future trends will involve the use of discriminative and
tools such as PCA and PLS reduce dimensionality of the predictive metabolomics as the ultimate tool for quality
data by linear combination of the original variables. In con- control. The metabolite profile of products meeting mini-
trast, random forest (RF) analysis permits multivariate data mum standards can be used as a baseline for quality
562 J.M. Cevallos-Cevallos et al. / Trends in Food Science & Technology 20 (2009) 557e566

acceptance. Individual samples obtained during processing peaks (Chen et al., 2007). Even though no attempt to deter-
can be analyzed and compared to the baseline through mine the limit of detection of E. coli was made, the technique
MVDA techniques to determine acceptability of the batch clearly shows potential for rapid pathogen detection in food.
produced. Moreover, accidental adulteration of food (e.g. Additionally, the same technique discriminated spoiled fish
allergenic inclusion or microbial contamination) can be de- through the presence of putrescine, cadaverine, and hista-
tected by appearance of uncommon peaks in the sample’s mine, showing a great potential of this type of analysis in
metabolic profile. Informative metabolomics can elucidate food safety. Informative and predictive metabolomics in
the nature of the peaks of interest. In addition, combina- fresh raw fish have been recommended as tools to provide ev-
tions of predictive and informative metabolomics have the idence of water contamination, temperature stress, and the
potential to become the single all-parameter analysis tool. fish health conditions at the moment of the catch (Samuels-
Quality parameters are usually individually measured com- son & Larsson, 2008).
plicated and costly protocols. Many of these parameters can Metabolomics has the potential to assess the safety of
be quantified in a single run of informative metabolomics novel pre- and post-harvest technologies. Unintended ef-
whereas others (e.g. sensory attributes) can be estimated fects of genetic modification of foods can been assessed
by predictive models based on sample metabolite profile, by untargeted discriminative analyses (Chao & Krewski,
providing a cost-effective alternative to quality analyses. 2008; Zdunczyk, 2006). Catchpole et al. (2005) utilized un-
Predictive models have been developed to estimate sensory targeted discriminative metabolomics to differentiate genet-
attributes of green tea (Ikeda et al., 2007; Pongsuwan et al., ically modified (GM) potatoes from non-treated ones.
2008; Tarachiwin et al., 2007), watermelon (Tarachiwin Sample differentiation occurred based on the intended var-
et al., 2008) and mushrooms (Cho et al., 2007). Similarly, iation of fructans in GM samples. After removal of fructans
metabolomics has the potential of identifying compounds derivatives from the model, no discrimination was observed,
that dictate consumer taste preferences. Consumers’ prefer- suggesting that GM potatoes are similar in composition to
ences can be obtained by taste panels while discriminating the original ones. Similarly, intended increase in flavonoid
compounds can be identified by metabolomics techniques. concentration in GM tomatoes have been confirmed
Sensory evaluations with various concentrations of the cho- through targeted informative metabolomics (Le Gall, Du-
sen compounds will confirm their impact on consumer pref- Pont, et al., 2003) whereas small non-intended variations
erences (Fig. 3). were detected by untargeted analysis (Le Gall, Colquhoun,
Davis, Collins, & Verhoeyen, 2003), concluding that no
Metabolomics in food safety major unintended changes occurred after genetic modifica-
Many untargeted discriminative tools have been applied tion. Future trends would involve the use of informative
in food safety. Amongst the many techniques, neutral desorp- metabolomics to assess the safety of new or controversial
tion extractive electrospray-ionization MS (EESI-MS) was processing technologies such as irradiation.
able to discriminate Escherichia coli-contaminated spinach
through the presence of unidentified high molecular weight Metabolomics for compliance with food regulations
Differences in food metabolite profile can be due among
other things to genotype and growing conditions (e.g. cli-
Food samples
mate, soil composition, fertilization, and irrigation). Estab-
lishing baseline regional and varietal variability in
metabolite profiles provides reliable information about the
Sensory analysis Metabolomics origin and authenticity of the food. Country of origin regu-
Discriminative and lations was verified by discriminative and predictive metab-
Taste panel informative
olomics. Origins of honey (Donarski et al., 2008), olive oil
(Cavaliere et al., 2007) and wine (Son et al., 2008), have
Discriminating
been determined by discriminative and predictive metabo-
Preferences lomics. Regulations in many countries do not allow
compounds
the use of GM foods, and GM verification analyses are
complicated and expensive. Discriminative and predictive
analysis have been used to differentiate genetic modifica-
New formulations tion in maize (Levandi, Leon, Kaljurand, Garcia-Canas,
& Cifuentes, 2008), soybean (Garcia-Villalba et al.,
2008), potatoes (Catchpole et al., 2005; Le Gall, Colqu-
Sensory houn, et al., 2003), and wheat (Shewry, Baudo, Lovegrove,
confirmation & Powers, 2007). Additionally, metabolomics can be used
for compliance verification of labeled ingredients. These
Fig. 3. Potential application of metabolomics for understanding con- analyses have relied in the use of discriminative metabolo-
sumer preferences. mics to differentiate among varieties of several fruits and
 J.M. Cevallos-Cevallos et al. / Trends in Food Science & Technology 20 (2009) 557e566 563

vegetables. For instance, MVDA techniques were used to incorporated into an algorithm that predicts the CFU.
differentiate cherry tomato from other varieties such as Metabolomics studies during E. coli growth have shown
beef and round tomatoes by SPMEeGCeMS (Tikunov the time-related progression of several metabolites (Koek,
et al., 2005), LCeMS and NMR (Moco, Forshed, De Muilwijk, & van der Werf, 2006).
Vos, Bino, & Vervoort, 2008). Variety differentiation has Moreover, metabolomics has the potential to find new
also been applied to broccoli (Luthria et al., 2008), wines antimicrobial compounds as well as to determine the analy-
(Pereira et al., 2007; Son et al., 2008), ginseng (Kang tes responsible for the antimicrobial characteristics of cer-
et al., 2008; Shin et al., 2007), and potatoes (Dobson tain plants and foods. Zhi, Yu, and Yi (2008) utilized
et al., 2008; Parr, Mellon, Colquhoun, & Davies, 2005). discriminative metabolomics based on HPLC to identify
dihydrocucurbitacin F-25-O-acetate as the major antimicro-
Metabolomics in food microbiology bial component of the herb Hemsleya pengxianensis. PCA
Research aimed at identifying bacterial contamination of data showed that Staphylococcus aureus treated with dihy-
foods has benefited from the use of metabolomics. Bacteria drocucurbitacin F-25-O-acetate clustered with those treated
identification and confirmation is traditionally done by with the herb extract.
complex numerous biochemical tests. In contrast, discrim-
inative and predictive analyses have the potential for iden- Metabolomics in food processing
tifying and confirming bacterial contamination. These Food processing involves the combination of physical and
analyses are mostly MS-based (Ecker et al., 2008). Micro- chemical events that may cause important changes in food
organisms are initially grown in culture media, then con- components. These changes can be detected by metabolo-
centrated (typically by centrifugation) and internal mics. The production of cheonggukjang (a soybean and rice
metabolites are extracted through cell disruption processes straw fermented drink) has been monitored by informative
such as ultrasound or bead beating processes before separa- and discriminative untargeted analysis using NMR (Choi,
tion or detection occurs. By following this method and the Yoon, Kim, & Kwon, 2007). The method showed the
use of a matrix assisted laser desorption/ionization time of expected time-related reduction of sugars (e.g. sucrose and
flight mass spectrometry (MALDIeTOFeMS) to detect fructose) and increase of acetic acid, tyrosine, phenylalanine
high molecular weight compounds, 12 species of Aspergil- and others. Final products were differentiated as a function of
lus and 5 strains of Aspergillus flavus have been classified fermentation time by PCA. In addition, Capanoglu et al.
with 95e100% accuracy (Hettick et al., 2008). Similar (2008) utilized both targeted and untargeted informative me-
methods have been used to classify E. coli and Yersinia tabolomic analyses to show that several flavonoids such as ru-
according to growing culture media, species, and strain tin, naringenin and derivates, as well as some alkaloids
(Parisi et al., 2008). Metabolomics can also be used for un- increased significantly after the breaking step (fruit chop-
derstanding microbial metabolism. Dynamics of glycolysis ping). The appearance of new flavonoids was explained by
in E. coli have been assessed under systemic variation of the possible activation of pertinent enzymes after wounding.
growth rate and different glucose availability (Schaub & In addition, reduction of these compounds after the pulping
Reuss, 2008) generating information on how glycolysis is step was observed because of the high presence of these ana-
affected under these conditions. lytes in the removed skin and seeds. Metabolomics can also
Wine and baking yeasts have been differentiated from be used to understand the suitability of certain varieties for
medical strains by using DIMS and GCeTOFeMS (MacK- processing. For instance, several potato varieties are preferred
enzie et al., 2008). In addition, exo-metabolites of several for frying whereas other for baking. To assess differences, po-
wine yeast strains were analyzed by HPLC and GCeFID tato varieties have been analyzed by flow infusion electro-
to compare aroma relevant compounds to gene expression spray-ionization MS (FI-ESI-MS) and compound identification
(Rossouw, Nas, & Bauer, 2008) showing the potential of was aided by GCeMS (Beckmann et al., 2007). Cultivars
metabolomics for assessing gene expressions. Salara and Agria were low in tyrosine (major substrate for
Current methods for quantification of bacteria in food polyphenol oxidases) making them suitable for slicing and
still rely on lengthy techniques such as plate count and frying. Tyrosine is also a precursor of aroma and flavor com-
most probable number. Metabolomic analysis coupled to pounds in boiled potatoes by Strecker degradation. Cultivars
sensor development can provide critical information for found to be high in tyrosine (Désirée and Granola) are more
the detection and quantification of bacteria. Metabolomics suitable for baking (Beckmann et al., 2007). This type of
has been successfully used for the discovery of specific bio- analysis has shown the potential for providing valuable infor-
markers in areas such as plant physiology (Glauser et al., mation to food product and process development industry.
2008). The discovery of bacteria biomarkers and their mon- Informative metabolomics has the potential to assess un-
itoring throughout growth phases has the potential to intended effects during processing and pre-processing such
become a quantitative tool if related to the final bacteria as changes in nutrient composition, degradation of health-
colony forming units (CFU). Sensors may be developed related compounds, and formation of new compounds like
to monitor the formation of the biomarker in the culture toxins. In addition informative and discriminative metabo-
broth and the rate of biomarker production can be lomics have the potential to study other pre-processing
564 J.M. Cevallos-Cevallos et al. / Trends in Food Science & Technology 20 (2009) 557e566

scenarios such as organic food production, and denomina- Bedair, M., & Sumner, L. W. (2008). Current and emerging mass-
tions such as ‘‘cage free’’ or ‘‘grade A’’. spectrometry technologies for metabolomics. Trends in Analytical
Chemistry, 27(3), 238e250.
Capanoglu, E., Beekwilder, J., Boyacioglu, D., Hall, R., & De Vos, R.
Conclusions (2008). Changes in antioxidant and metabolite profiles during
Metabolomics has shown to be an important tool for the production of tomato paste. Journal of Agricultural and Food
progress of the main food science areas such as compliance Chemistry, 56, 964e973.
Catchpole, G. S., Beckmann, M., Enot, D. P., Mondhe, M., Zywicki, B.,
of regulations, processing, quality, safety, and microbiol-
Taylor, J., et al. (2005). Hierarchical metabolomics demonstrates
ogy. Recent studies suggest that the potential of metabolo- substantial compositional similarity between genetically modified
mics in food science can be expanded to the area of food and conventional potato crops. Proceedings of the National
product development by determining the compounds re- Academy of Sciences of the United States of America, 102(40),
sponsible for consumers’ taste preferences. 14458e14462.
Cavaliere, B., De Nino, A., Hayet, F., Lazez, A., Macchione, B.,
The development of rapid technologies such as DIMS,
Moncef, C., et al. (2007). A metabolomic approach to the evalu-
IMS, and EESI has helped the growth of metabolomics in ation of the origin of extra virgin olive oil: a convenient statistical
food science. However, further improvement on these tech- treatment of mass spectrometric analytical data. Journal of
niques is necessary to overcome sensitivity and compound Agricultural and Food Chemistry, 55(4), 1454e1462.
identification issues. Cevallos-Cevallos, J., Rouseff, R., & Reyes-De-Corcuera, J. (2009). Un-
targeted metabolite analysis of healthy and Huanglongbing infected
Most of the data analyses in food have relied in linear
orange leaves by CEeDAD. Electrophoresis, 30(7), 1240e1248.
MVDA tools, not considering possible non-linear aspects Chao, E., & Krewski, D. (2008). A risk-based classification scheme for
of the samples. Future trends should involve the use of genetically modified foods II: graded testing. Regulatory Toxicol-
non-linear tools for dimensionality reduction in food ogy and Pharmacology, 52(3), 223e234.
metabolomics. Chen, H. W., Wortmann, A., & Zenobi, R. (2007). Neutral desorption
sampling coupled to extractive electrospray ionization mass
Even though metabolomic analyses in food have been
spectrometry for rapid differentiation of bilosamples by metabo-
much diversified, most studies can be considered as dis- lomic fingerprinting. Journal of Mass Spectrometry, 42(9),
criminative (Table 1) with very few compounds identified. 1123e1135.
Therefore, the development of a food metabolome data- Cho, I. H., Kim, Y. S., & Choi, H. K. (2007). Metabolomic discrimi-
base, as suggested by Wishart (2008b) is needed in order nation of different grades of pine-mushroom (Tricholoma matsu-
take Sing.) using H-1 NMR spectrometry and multivariate data
to facilitate compound identification and the development analysis. Journal of Pharmaceutical and Biomedical Analysis,
of informative metabolomics. In addition, most reports 43(3), 900e904.
have focused on fruit and vegetables (Table 1) leaving the Choi, H. K., Yoon, J. H., Kim, Y. S., & Kwon, D. Y. (2007). Metabolomic
meat, seafood, and related areas still underexplored. Be- profiling of Cheonggukjang during fermentation by H-1 NMR
cause of some metabolic similarities, identification of spectrometry and principal components analysis. Process
Biochemistry, 42(2), 263e266.
many compounds in red meat can be carried out by using
Cozzolino, D., Flood, L., Bellon, J., Gishen, M., & Lopes, M. D. (2006).
available human metabolome databases (Wishart, 2008b). Combining near infrared spectroscopy and multivariate analysis as
Metabolomics’ successful association to other analytical a tool to differentiate different strains of Saccharomyces cerevisiae:
areas such as genomics has been demonstrated, showing a metabolomic study. Yeast, 23(14e15), 1089e1096.
the potential of metabolite profiling to be linked to other Defernez, M., Gunning, Y. M., Parr, A. J., Shepherd, L. V. T.,
Davies, H. V., & Colquhoun, I. J. (2004). NMR and HPLCeUV
areas as well. Metabolomics as a first step for sensor devel- profiling of potatoes with genetic modifications to metabolic
opment has the potential to introduce a series of new rapid pathways. Journal of Agricultural and Food Chemistry, 52(20),
methods for food analysis. In this area, bacteria biomarkers 6075e6085.
can be discovered by metabolomics techniques and sensors Dixon, R. A., Gang, D. R., Charlton, A. J., Fiehn, O., Kuiper, H. A.,
can be developed for rapid detection of the selected bio- Reynolds, T. L., et al. (2006). Perspective e applications of me-
tabolomics in agriculture. Journal of Agricultural and Food
markers. To achieve this purpose, studies on microbial bio- Chemistry, 54, 8984e8994.
markers identification involving different levels of bacterial Dobson, G., Shepherd, T., Verrall, S. R., Conner, S., McNicol, J. W.,
contamination, accompanying flora, and biomarker response Ramsay, G., et al. (2008). Phytochemical diversity in tubers of
in food are needed. potato cultivars and landraces using a GCeMS metabolomics
The rapid growth of metabolomics in the food science approach. Journal of Agricultural and Food Chemistry, 56(21),
10280e10291.
area suggests the potential of discriminative, predictive, Donarski, J. A., Jones, S. A., & Charlton, A. J. (2008). Application of
and informative analyses to solve the most important prob- cryoprobe H-1 nuclear magnetic resonance spectroscopy and
lems and provide important information to the food industry. multivariate analysis for the verification of Corsican honey. Journal
of Agricultural and Food Chemistry, 56(14), 5451e5456.
Ecker, D. J., Sampath, R., Massire, C., Blyn, L. B., Hall, T. A.,
References
Eshoo, M. W., et al. (2008). Innovation e Ibis T5000: a universal
Beckmann, M., Enot, D. P., Overy, D. P., & Draper, J. (2007). Repre- biosensor approach for microbiology. Nature Reviews Microbio-
sentation, comparison, and interpretation of metabolome finger- logy, 6(7), 553e558.
print data for total composition analysis and quality trait Garcia-Villalba, R., Leon, C., Dinelli, G., Segura-Carretero, A.,
investigation in potato cultivars. Journal of Agricultural and Food Fernandez-Gutierrez, A., Garcia-Canas, V., et al. (2008). Compar-
Chemistry, 55(9), 3444e3451. ative metabolomic study of transgenic versus conventional soybean
 J.M. Cevallos-Cevallos et al. / Trends in Food Science & Technology 20 (2009) 557e566 565

using capillary electrophoresis-time-of-flight mass spectrometry. flavonoid glycosides in genetically modified tomato (Lycopersicon
Journal of Chromatography A, 1195(1e2), 164e173. esculentum) fruits. Journal of Agricultural and Food Chemistry,
Glauser, G., Guillarme, D., Grata, E., Boccard, J., Thiocone, A., 51(9), 2438e2446.
Carrupt, P.-A., et al. (2008). Optimized liquid chromatographye Levandi, T., Leon, C., Kaljurand, M., Garcia-Canas, V., & Cifuentes, A.
mass spectrometry approach for the isolation of minor stress (2008). Capillary electrophoresis time-of-flight mass spectrometry
biomarkers in plant extracts and their identification by capillary for comparative metabolomics of transgenic versus conventional
nuclear magnetic resonance. Journal of Chromatography A, maize. Analytical Chemistry, 80(16), 6329e6335.
1180(1e2), 90e98. Liu, C., Si, D., Wan, R., Lin, Y., & Xu, Y. (2008). Metabonomics in
Green, R. D., Qureshi, M. A., Long, J. A., Burfening, P. J., & research of natural drugs and traditional Chinese medicines. Chi-
Hamernik, D. L. (2007). Identifying the future needs for long-term nese Journal of Natural Medicines, 6(2), 81e88.
USDA efforts in agricultural animal genomics. International Journal Luthria, D. L., Lin, L. Z., Robbins, R. J., Finley, J. W., Banuelos, G. S., &
of Biological Sciences, 3(3), 185e191. Harnly, J. M. (2008). Discriminating between cultivars and treat-
Gullberg, J., Jonsson, P., Nordstrom, A., Sjostrom, M., & Moritz, T. ments of broccoli using mass spectral fingerprinting and analysis of
(2004). Design of experiments: an efficient strategy to identify variance-principal component analysis. Journal of Agricultural and
factors influencing extraction and derivatization of Arabidopsis Food Chemistry, 56(21), 9819e9827.
thaliana samples in metabolomic studies with gas chromatography/ Ma, C., Wang, H., Lu, X., Xu, G., & Liu, B. (2008). Metabolic finger-
mass spectrometry. Analytical Biochemistry, 331(2), 283e295. printing investigation of Artemisia annua L. in different stages of
Hall, R. D. (2007). Food metabolomics: META-PHOR e a new development by gas chromatography and gas chromatographye
European research initiative. Agro Food Industry Hi-Tech, 18(1), mass spectrometry. Journal of Chromatography A, 1186(1e2),
14e16. 412e419.
Hall, R. D., Brouwer, I. D., & Fitzgerald, M. A. (2008). Plant metabo- McDougall, G., Martinussen, I., & Stewart, D. (2008). Towards fruitful
lomics and its potential application for human nutrition. Physio- metabolomics: high throughput analyses of polyphenol composi-
logia Plantarum, 132(2), 162e175. tion in berries using direct infusion mass spectrometry. Journal of
Hettick, J. M., Green, B. J., Buskirk, A. D., Kashon, M. L., Slaven, J. E., Chromatography B e Analytical Technologies in the Biomedical
Janotka, E., et al. (2008). Discrimination of Aspergillus isolates at and Life Sciences, 871(2), 362e369.
the species and strain level by matrix-assisted laser desorption/ MacKenzie, D. A., Defernez, M., Dunn, W. B., Brown, M., Fuller, L. J.,
ionization time-of-flight mass spectrometry fingerprinting. Analyt- de Herrera, S., et al. (2008). Relatedness of medically important
ical Biochemistry, 380(2), 276e281. strains of Saccharomyces cerevisiae as revealed by phylogenetics
Ikeda, T., Kanaya, S., Yonetani, T., Kobayashi, A., & Fukusaki, E. and metabolomics. Yeast, 25(7), 501e512.
(2007). Prediction of Japanese green tea ranking by fourier trans- Moalemiyan, M., Vikram, A., & Kushalappa, A. C. (2007). Detection
form near-infrared reflectance spectroscopy. Journal of Agricultural and discrimination of two fungal diseases of mango (cv. Keitt) fruits
and Food Chemistry, 55, 9908e9912. based on volatile metabolite profiles using GC/MS. Postharvest
Jahangir, M., Kim, H. K., Choi, Y. H., & Verpoorte, R. (2008). Metab- Biology and Technology, 45(1), 117e125.
olomic response of Brassica rapa submitted to pre-harvest bacterial Moco, S., Forshed, J., De Vos, R. C. H., Bino, R. J., & Vervoort, J.
contamination. Food Chemistry, 107, 362e368. (2008). Intra- and inter-metabolite correlation spectroscopy of
Kaddurah-Daouk, R., & Krishnan, K. R. R. (2009). Metabolomics: tomato metabolomics data obtained by liquid chromatographye
a global biochemical approach to the study of central nervous mass spectrometry and nuclear magnetic resonance. Metabolo-
system diseases. Neuropsychopharmacology, 34(1), 173e186. mics, 4(3), 202e215.
Kang, J., Lee, S., Kang, S., Kwon, H., Park, J., Kwon, S., et al. (2008). Monton, M. R. N., & Soga, T. (2007). Metabolome analysis by capillary
NMR-based metabolomics approach for the differentiation of gin- electrophoresisemass spectrometry. Journal of Chromatography A,
seng (Panax ginseng) roots from different origins. Archives of 1168(1e2), 237e246.
Pharmacal Research, 31(3), 330e336. Parisi, D., Magliulo, M., Nanni, P., Casale, M., Forina, M., & Roda, A.
Kemsley, E. K., Le Gall, G., Dainty, J. R., Watson, A. D., Harvey, L. J., (2008). Analysis and classification of bacteria by matrix-assisted
Tapp, H. S., et al. (2007). Multivariate techniques and their appli- laser desorption/ionization time-of-flight mass spectrometry and
cation in nutrition: a metabolomics case study. British Journal of a chemometric approach. Analytical and Bioanalytical Chemistry,
Nutrition, 98(1), 1e14. 391(6), 2127e2134.
Koek, M. M., Muilwijk, B., & van der Werf, M. J. (2006). Microbial Parr, A. J., Mellon, F. A., Colquhoun, I. J., & Davies, H. V. (2005).
metabolomics with gas chromatography/mass spectrometry. Ana- Dihydrocaffeoyl polyamines (kukoamine and allies) in potato
lytical Chemistry, 78(4), 1272e1281. (Solanum tuberosum) tubers detected during metabolite profiling.
Kooy, F.v. d., Verpoorte, R., & Meyer, J. J. M. (2008). Metabolomic Journal of Agricultural and Food Chemistry, 53(13), 5461e5466.
quality control of claimed anti-malarial Artemisia afra herbal Pereira, G. E., Gaudillere, J. P., Van Leeuwen, C., Hilbert, G.,
remedy and A. afra and A. annua plant extracts. South African Maucourt, M., Deborde, C., et al. (2007). H-1-NMR metabolic
Journal of Botany, 74(2), 186e189. profiling of wines from three cultivars, three soil types and two
Kushalappa, A. C., Vikram, A., & Raghavan, G. S. V. (2008). Metabo- contrasting vintages. Journal International Des Sciences De La
lomics of headspace gas for diagnosing diseases of fruits and veg- Vigne Et Du Vin, 41, 103e109.
etables after harvest. Stewart Postharvest Review, 4(2). article10. Pongsuwan, W., Bamba, T., Harada, K., Yonetani, T., Kobayashi, A., &
Lee, G., Rodriguez, C., & Madabhushi, A. (2008). Investigating the Fukusaki, E. (2008). High-throughput technique for comprehensive
efficacy of nonlinear dimensionality reduction schemes in classi- analysis of Japanese green tea quality assessment using ultra-per-
fying gene and protein expression studies. IEEE-ACM Transactions formance liquid chromatography with time-of-flight mass spec-
on Computational Biology and Bioinformatics, 5(3), 368e384. trometry (UPLC/TOF MS). Journal of Agricultural and Food
Le Gall, G., Colquhoun, I. J., Davis, A. L., Collins, G. J., & Chemistry, 56(22), 10705e10708.
Verhoeyen, M. E. (2003). Metabolite profiling of tomato (Lycoper- Ramautar, R., Demirci, A., & Jong, G. J.d. (2006). Capillary electro-
sicon esculentum) using H-1 NMR spectroscopy as a tool to detect phoresis in metabolomics. Trends in Analytical Chemistry,
potential unintended effects following a genetic modification. 25(5), 12.
Journal of Agricultural and Food Chemistry, 51(9), 2447e2456. Rochfort, S. (2005). Metabolomics reviewed: a new ‘‘Omics’’ platform
Le Gall, G., DuPont, M. S., Mellon, F. A., Davis, A. L., Collins, G. J., technology for systems biology and implications for natural prod-
Verhoeyen, M. E., et al. (2003). Characterization and content of ucts research. Journal of Natural Products, 68(12), 1813e1820.
566 J.M. Cevallos-Cevallos et al. / Trends in Food Science & Technology 20 (2009) 557e566

Rochfort, S. J., Trenerry, V. C., Imsic, M., Panozzo, J., & Jones, R. analysis for metabolomics. Large-scale profiling of tomato fruit vol-
(2008). Class targeted metabolomics: ESI ion trap screening atiles. Plant Physiology, 139(3), 1125e1137.
methods for glucosinolates based on MSn fragmentation. Phyto- Toyo’oka, T. (2008). Determination methods for biologically active
chemistry, 69(8), 1671e1679. compounds by ultra-performance liquid chromatography coupled
Rossouw, D., Nas, T., & Bauer, F. F. (2008). Linking gene regulation with mass spectrometry: application to the analyses of pharma-
and the exo-metabolome: a comparative transcriptomics approach ceuticals, foods, plants, environments, metabonomics, and me-
to identify genes that impact on the production of volatile aroma tabolomics. Journal of Chromatographic Science, 46(3), 233e247.
compounds in yeast. BMC Genomics, 9(530). Ursem, R., Tikunov, Y., Bovy, A., van Berloo, R., & van Eeuwijk, F.
Rudell, D. R., Mattheis, J. P., & Curry, F. A. (2008). Prestorage (2008). A correlation network approach to metabolic data analysis
ultraviolet-white light irradation alters apple peel metabolome. for tomato fruits. Euphytica, 161(1e2), 181e193.
Journal of Agricultural and Food Chemistry, 56(3), 1138e1147. Vautz, W., Zimmermann, D., Hartmann, M., Baumbach, J. I., Nolte, J.,
Samuelsson, L. M., & Larsson, D. G. J. (2008). Contributions from & Jung, J. (2006). Ion mobility spectrometry for food quality and
metabolomics to fish research. Molecular Biosystems, 4(10), safety. Food Additives and Contaminants, 23(11), 1064e1073.
974e979. Vikram, A., Hamzehzarghani, H., & Kushalappa, A. C. (2005). Volatile
Schaub, J., & Reuss, M. (2008). In vivo dynamics of glycolysis in metabolites from the headspace of onion bulbs inoculated with
Escherichia coli shows need for growth-rate dependent metabo- postharvest pathogens as a tool for disease discrimination. Revue
lome analysis. Biotechnology Progress, 24(6), 1402e1407. Canadienne De Phytopathologie. [Canadian Journal of Plant
Shewry, P. R., Baudo, M., Lovegrove, A., & Powers, S. (2007). Are GM Pathology], 27(2), 194e203.
and conventionally bred cereals really different? Trends in Food Vikram, A., Lui, L. H., Hossain, A., & Kushalappa, A. C. (2006). Met-
Science & Technology, 18(4), 201e209. abolic fingerprinting to discriminate diseases of stored carrots.
Shin, Y. S., Bang, K. H., In, D. S., Kim, O. T., Hyun, D. Y., Ahn, I. O., et Annals of Applied Biology, 148(1), 17e26.
al. (2007). Fingerprinting analysis of fresh ginseng roots of different Vikram, A., Prithiviraj, B., Hamzehzarghani, H., & Kushalappa, A.
ages using H-1-NMR spectroscopy and principal components (2004). Volatile metabolite profiling to discriminate diseases of
analysis. Archives of Pharmacal Research, 30, 1625e1628. McIntosh apple inoculated with fungal pathogens. Journal of the
Son, H. S., Kim, K. M., Van den Berg, F., Hwang, G. S., Park, W. M., Science of Food and Agriculture, 84(11), 1333e1340.
Lee, C. H., et al. (2008). H-1 nuclear magnetic resonance-based van der Werf, M. J., Jellema, R. H., & Hankemeier, T. (2005). Microbial
metabolomic characterization of wines by grape varieties and metabolomics: replacing trial-and-error by the unbiased selection
production areas. Journal of Agricultural and Food Chemistry, and ranking of targets. Journal of Industrial Microbiology
56(17), 8007e8016. & Biotechnology, 32(6), 234e252.
Stewart, D., McDougall, G. J., Sungurtas, J., Verrall, S., Graham, J., & Wishart, D. S. (2008a). Applications of metabolomics in drug discov-
Martinussen, I. (2007). Metabolomic approach to identifying bio- ery and development. Drugs in R&D, 9(5), 307e322.
active compounds in berries: advances toward fruit nutritional Wishart, D. S. (2008b). Metabolomics: applications to food science
enhancement. Molecular Nutrition & Food Research, 51(6), and nutrition research. Trends in Food Science & Technology,
645e651. 19(9), 482e493.
Sumner, L. W., Urbanczyk-Wochniak, E., & Broeckling, C. D. (2008). Wishart, D. S., Tzur, D., Knox, C., Eisner, R., Guo, A. C., Young, N.,
Metabolomics data analysis, visualization, and integration. In: et al. (2007). HMDB: the human metabolome database. Nucleic
Plant bioinformatics (pp. 409e436). Acids Research, 35, D521eD526.
Tarachiwin, L., Masako, O., & Fukusaki, E. (2008). Quality evaluation Yin, H. (2007). Nonlinear dimensionality reduction and data visuali-
and prediction of Citrullus lanatus by H-1 NMR-based metabolo- zation: a review. International Journal of Automation and Com-
mics and multivariate analysis. Journal of Agricultural and Food puting, 4(3), 294e303.
Chemistry, 56(14), 5827e5835. Zdunczyk, Z. (2006). New bioanalytical technologies (‘‘Omics’’) in the
Tarachiwin, L., Ute, K., Kobayashi, A., & Fukusakii, E. (2007). H-1 evaluation of biological properties of foods and feeds. Polish
NMR based metabolic profiling in the evaluation of Japanese green Journal of Natural Sciences, Suppl. 3, 33e38.
tea quality. Journal of Agricultural and Food Chemistry, 55(23), Zhi, B. Y., Yu, Y., & Yi, Z. L. (2008). Investigation of antimicrobial
9330e9336. model of Hemsleya pengxianensis W.J. Chang and its main active
Tikunov, Y., Lommen, A., de Vos, C. H. R., Verhoeven, H. A., Bino, R. J., component by metabolomics technique. Journal of Ethnopharma-
Hall, R. D., et al. (2005). A novel approach for non-targeted data cology, 116(1), 89e95.