Metabolic Phenotyping and Systems Biology Approaches to Understanding Metabolic Syndrome and Fatty Liver Disease

Metabolic syndrome, a cluster of risk factors for type 2 diabetes mellitus and cardiovascular disease, is becoming an increasing global health concern. Insulin resistance is often associated with metabolic syndrome and also typical hepatic manifestations such as nonalcoholic fatty liver disease. Profiling of metabolic products (metabolic phenotyping or metabotyping) has provided new insights into metabolic syndrome and nonalcoholic fatty liver disease. Data from nuclear magnetic resonance spectroscopy and mass spectrometry combined with statistical modeling and top-down systems biology have allowed us to analyze and interpret metabolic signatures in terms of metabolic pathways and protein interaction networks and to identify the genomic and metagenomic determinants of metabolism. For example, metabolic phenotyping has shown that relationships between host cells and the microbiome affect development of the metabolic syndrome and fatty liver disease. We review recent developments in metabolic phenotyping and systems biology technologies and how these methodologies have provided insights into the mechanisms of metabolic syndrome and nonalcoholic fatty liver disease. We discuss emerging areas of research in this field and outline our vision for how metabolic phenotyping could be used to study metabolic syndrome and fatty liver disease. Metabolic syndrome, a cluster of risk factors for type 2 diabetes mellitus and cardiovascular disease, is becoming an increasing global health concern. Insulin resistance is often associated with metabolic syndrome and also typical hepatic manifestations such as nonalcoholic fatty liver disease. Profiling of metabolic products (metabolic phenotyping or metabotyping) has provided new insights into metabolic syndrome and nonalcoholic fatty liver disease. Data from nuclear magnetic resonance spectroscopy and mass spectrometry combined with statistical modeling and top-down systems biology have allowed us to analyze and interpret metabolic signatures in terms of metabolic pathways and protein interaction networks and to identify the genomic and metagenomic determinants of metabolism. For example, metabolic phenotyping has shown that relationships between host cells and the microbiome affect development of the metabolic syndrome and fatty liver disease. We review recent developments in metabolic phenotyping and systems biology technologies and how these methodologies have provided insights into the mechanisms of metabolic syndrome and nonalcoholic fatty liver disease. We discuss emerging areas of research in this field and outline our vision for how metabolic phenotyping could be used to study metabolic syndrome and fatty liver disease. Metabolic syndrome is a cluster of disorders that increase the risk of type 2 diabetes mellitus and cardiovascular disease as defined in a joint consensus statement from international cardiovascular and diabetes groups.1Alberti K.G. Eckel R.H. Grundy S.M. et al.Harmonizing the metabolic syndrome: a joint interim statement of the International Diabetes Federation Task Force on Epidemiology and Prevention; National Heart, Lung, and Blood Institute; American Heart Association; World Heart Federation; International Atherosclerosis Society; and International Association for the Study of Obesity.Circulation. 2009; 120: 1640-1645Crossref PubMed Scopus (8476) Google Scholar This syndrome is associated with obesity and leads to disease in numerous organ systems, so organ-specific features are unlikely to provide a definitive phenotype. The main hepatic phenotypic manifestation of insulin resistance is primary nonalcoholic fatty liver disease (NAFLD). Approximately 34% of Western adults older than 20 years of age have metabolic syndrome2Ervin R.B. Prevalence of metabolic syndrome among adults 20 years of age and over, by sex, age, race and ethnicity, and body mass index: United States, 2003-2006.Natl Health Stat Rep. 2009; : 1-7PubMed Google Scholar, 3Eckel R.H. Alberti K.G. Grundy S.M. et al.The metabolic syndrome.Lancet. 2010; 375: 181-183Abstract Full Text Full Text PDF PubMed Scopus (673) Google Scholar and 33.7% of US adults are obese (body mass index >30 kg/m2), with associated and increasing rates of type 2 diabetes.4Roger V.L. Go A.S. Lloyd-Jones D.M. et al.Heart disease and stroke statistics—2011 update: a report from the American Heart Association.Circulation. 2011; 123: e18-e209Crossref PubMed Scopus (4068) Google Scholar Many genomic and environmental factors affect the risk of developing metabolic syndrome, including age, sex, ethnicity, socioeconomic status, smoking status, diabetes, and amount of exercise.4Roger V.L. Go A.S. Lloyd-Jones D.M. et al.Heart disease and stroke statistics—2011 update: a report from the American Heart Association.Circulation. 2011; 123: e18-e209Crossref PubMed Scopus (4068) Google Scholar Also, the cardiovascular risk associated with metabolic syndrome varies based on the combination of the components of metabolic syndrome, whereas NAFLD remains an independent factor.5Franco O.H. Massaro J.M. Civil J. et al.Trajectories of entering the metabolic syndrome: the Framingham Heart Study.Circulation. 2009; 120: 1943-1950Crossref PubMed Scopus (115) Google Scholar, 6Gami A.S. Witt B.J. Howard D.E. et al.Metabolic syndrome and risk of incident cardiovascular events and death: a systematic review and meta-analysis of longitudinal studies.J Am Coll Cardiol. 2007; 49: 403-414Crossref PubMed Scopus (1337) Google ScholarThe spectrum of fatty liver disease comprises NAFLD and nonalcoholic steatohepatitis (NASH) with and without fibrosis, cirrhosis, and hepatocellular carcinoma.7Michelotti G.A. Machado M.V. Diehl A.M. NAFLD, NASH and liver cancer.Nat Rev Gastroenterol Hepatol. 2013; 10: 656-665Crossref PubMed Scopus (566) Google Scholar NAFLD is defined histologically, based on the presence of simple hepatic steatosis without hepatocyte injury (ballooning), whereas NASH is defined by the presence of steatosis and inflammation with ballooning, with or without fibrosis.8Chalasani N. Younossi Z. Lavine J.E. et al.The diagnosis and management of non-alcoholic fatty liver disease: practice guideline by the American Gastroenterological Association, American Association for the Study of Liver Diseases, and American College of Gastroenterology.Gastroenterology. 2012; 142: 1592-1609Abstract Full Text Full Text PDF PubMed Scopus (1160) Google Scholar Metabolic syndrome is a strong predictor of the presence of steatohepatitis in patients with NAFLD,9Kang H. Greenson J.K. Omo J.T. et al.Metabolic syndrome is associated with greater histologic severity, higher carbohydrate, and lower fat diet in patients with NAFLD.Am J Gastroenterol. 2006; 101: 2247-2253Crossref PubMed Scopus (148) Google Scholar although the exact prevalence of NAFLD varies with the method of diagnosis; for instance, the prevalence of NAFLD can reach 46% when it is diagnosed by ultrasonography or magnetic resonance imaging, with or without measurement of hepatic enzyme levels.10Vernon G. Baranova A. Younossi Z.M. Systematic review: the epidemiology and natural history of non-alcoholic fatty liver disease and non-alcoholic steatohepatitis in adults.Aliment Pharmacol Ther. 2011; 34: 274-285Crossref PubMed Scopus (1980) Google Scholar, 11Williams C.D. Stengel J. Asike M.I. et al.Prevalence of nonalcoholic fatty liver disease and nonalcoholic steatohepatitis among a largely middle-aged population utilizing ultrasound and liver biopsy: a prospective study.Gastroenterology. 2011; 140: 124-131Abstract Full Text Full Text PDF PubMed Scopus (1385) Google Scholar The true incidence of NASH is likely to be underreported8Chalasani N. Younossi Z. Lavine J.E. et al.The diagnosis and management of non-alcoholic fatty liver disease: practice guideline by the American Gastroenterological Association, American Association for the Study of Liver Diseases, and American College of Gastroenterology.Gastroenterology. 2012; 142: 1592-1609Abstract Full Text Full Text PDF PubMed Scopus (1160) Google Scholar; in magnetic resonance spectroscopy studies without liver biopsies, the prevalence of NAFLD has been reported to be as high as 65% among obese patients.12Browning J.D. Szczepaniak L.S. Dobbins R. et al.Prevalence of hepatic steatosis in an urban population in the United States: impact of ethnicity.Hepatology. 2004; 40: 1387-1395Crossref PubMed Scopus (2723) Google Scholar Epidemiological studies have indicated that NAFLD13Ekstedt M. Franzén L.E. Mathiesen U.L. et al.Long-term follow-up of patients with NAFLD and elevated liver enzymes.Hepatology. 2006; 44: 865-873Crossref PubMed Scopus (1639) Google Scholar, 14Dam-Larsen S. Becker U. Franzmann M.-B. et al.Final results of a long-term, clinical follow-up in fatty liver patients.Scand J Gastroenterol. 2009; 44: 1236-1243Crossref PubMed Scopus (147) Google Scholar, 15Lazo M. Hernaez R. Bonekamp S. et al.Non-alcoholic fatty liver disease and mortality among US adults: prospective cohort study.BMJ. 2011; 343: d6891Crossref PubMed Scopus (244) Google Scholar has varying effects on mortality, depending on the diagnostic approach (NAFLD fibrosis score16Kim D. Kim W.R. Kim H.J. et al.Association between noninvasive fibrosis markers and mortality among adults with nonalcoholic fatty liver disease in the United States.Hepatology. 2013; 57: 1357-1365Crossref PubMed Scopus (421) Google Scholar, 17Treeprasertsuk S. Björnsson E. Enders F. et al.NAFLD fibrosis score: a prognostic predictor for mortality and liver complications among NAFLD patients.World J Gastroenterol. 2013; 19: 1219-1229Crossref PubMed Scopus (82) Google Scholar or disease severity). Irrespective of mortality, the morbidity and economic burden are significant and NAFLD is now the most common cause of liver disease in the West, accounting for an increasing proportion of patients who undergo liver transplantation (15%–20%).18Newsome P.N. Allison M.E. Andrews P.A. et al.Guidelines for liver transplantation for patients with non-alcoholic steatohepatitis.Gut. 2012; 61: 484-500Crossref PubMed Scopus (52) Google ScholarConsiderable challenges remain in the diagnosis, staging, and prognosis of metabolic syndrome, NAFLD, and NASH.19Jepsen P. Grønbæk H. Prognosis and staging of non-alcoholic fatty liver disease.BMJ. 2011; 343: d7302Crossref PubMed Scopus (0) Google Scholar For example, results from tests of liver biochemistry can be within normal ranges for patients with NAFLD and NASH, so these are not sufficiently sensitive for screening. Composite clinical scores that predict which patients will develop fibrosis, such as the NAFLD fibrosis score, lack specificity.20Wieckowska A. Zein N.N. Yerian L.M. et al.In vivo assessment of liver cell apoptosis as a novel biomarker of disease severity in nonalcoholic fatty liver disease.Hepatology. 2006; 44: 27-33Crossref PubMed Scopus (546) Google Scholar A definitive diagnosis requires liver biopsy analysis, which is expensive, is invasive, and has associated morbidities.19Jepsen P. Grønbæk H. Prognosis and staging of non-alcoholic fatty liver disease.BMJ. 2011; 343: d7302Crossref PubMed Scopus (0) Google ScholarA molecular definition of the syndrome would offer advantages over a phenotypic definition, because the disorder could be diagnosed at early stages, populations could be stratified based on risk, and therapeutic strategies could be selected for specific groups of people. This will likely require a multimodal “omics” approach that integrates multiple factors determining the metabolic phenotype in this complex and multifactorial disorder because obesity affects the cardiovascular system by causing insulin resistance,21Cusi K. Role of obesity and lipotoxicity in the development of nonalcoholic steatohepatitis: pathophysiology and clinical implications.Gastroenterology. 2012; 142: 711-725.e6Abstract Full Text Full Text PDF PubMed Scopus (528) Google Scholar although numerous pathways have been shown to be involved in the pathophysiology of metabolic syndrome.22Henao-Mejia J. Elinav E. Jin C. et al.Inflammasome-mediated dysbiosis regulates progression of NAFLD and obesity.Nature. 2012; 482: 179-185Crossref PubMed Scopus (1475) Google ScholarJust as we are able to analyze entire genomes and gene expression profiles, we are also able to characterize metabolic profiles of cells, tissues, and organisms via metabonomic and metabolomic analyses.23Nicholson J.K. Lindon J.C. Holmes E. “Metabonomics”: understanding the metabolic responses of living systems to pathophysiological stimuli via multivariate statistical analysis of biological NMR spectroscopic data.Xenobiotica. 1999; 29: 1181-1189Crossref PubMed Google Scholar, 24Nicholson J.K. Connelly J. Lindon J.C. et al.Metabonomics: a platform for studying drug toxicity and gene function.Nat Rev Drug Discov. 2002; 1: 153-161Crossref PubMed Scopus (1643) Google Scholar, 25Oliver S.G. Winson M.K. Kell D.B. et al.Systematic functional analysis of the yeast genome.Trends Biotechnol. 1998; 16: 373-378Abstract Full Text Full Text PDF PubMed Scopus (804) Google Scholar Metabolic profile analysis (metabolic phenotyping or metabotyping)26Gavaghan C.L. Holmes E. Lenz E. et al.An NMR-based metabonomic approach to investigate the biochemical consequences of genetic strain differences: application to the C57BL10J and Alpk:ApfCD mouse.FEBS Lett. 2000; 484: 169-174Crossref PubMed Scopus (285) Google Scholar can be used to identify specific features of metabolic syndrome. The metabolic profile results from genomic and environmental features of a cell or organism. Preclinical and clinical studies have generated unique insights into the molecular mechanisms of disease progression. Metabolic phenotyping can be used to identify biomarkers and different phenotypes of metabolic syndrome. We review recent technical developments in high-throughput metabotyping and top-down systems biology approaches (ie, data-driven approaches, as opposed to bottom-up model- and annotation-driven approaches), along with what we have learned about metabolic syndrome, NAFLD, and NASH via metabolic analyses. We also discuss the clinical applications that can be derived by increasing our understanding of the metabolic processes involved in metabolic syndrome and fatty liver disease.Metabolic Phenotyping and Systems BiologyUnlike transcripts and proteins, metabolites are not directly encoded by the genome (Table 1). The comprehensive measurement and analysis of metabolites and their variation in reaction to genetic or external stimuli is usually referred to as metabolomics25Oliver S.G. Winson M.K. Kell D.B. et al.Systematic functional analysis of the yeast genome.Trends Biotechnol. 1998; 16: 373-378Abstract Full Text Full Text PDF PubMed Scopus (804) Google Scholar and metabonomics,23Nicholson J.K. Lindon J.C. Holmes E. “Metabonomics”: understanding the metabolic responses of living systems to pathophysiological stimuli via multivariate statistical analysis of biological NMR spectroscopic data.Xenobiotica. 1999; 29: 1181-1189Crossref PubMed Google Scholar although both use the same experimental tools. Metabolic phenotyping in particular corresponds to the use of analytical chemistry methods in metabolomics to generate high-resolution metabolic observations about various disease and treatment conditions. Such characterization of molecular phenotypes in a biomedical environment can be extended to the term phenome. From an analytical point of view, a phenome is an integrated set of measureable physical and clinical features coupled to chemical, metabolic, and physiological properties that define biological subclasses. On a more philosophical level, it is the direct product of gene-environment (exposome) interactions on an individual or group operating throughout development and life: a dynamic property.Table 1High-Throughput “Omics” Approaches in Integrative BiologyTechnologyMoleculeKnowledgeLimitsClinical applicationsGenomicsDNAGenetic polymorphisms, haplotypes, full genome sequence through next-generation sequencingRestricted to genetic determinants, ignores the environmentGenome-wide association studiesPersonal genomicsTranscriptomicsRNAExpression patternsExpression does not necessarily match protein abundanceBlood and biopsy transcript signaturesProteomicsProteinsTargeted and untargeted protein abundance profilesProtein abundance does not mean functionProtein identification can be challengingNumerous clinical applications for measurement of circulating endogenous protein markersMetabonomics (understanding the response of living systems to stimuli23Nicholson J.K. Lindon J.C. Holmes E. “Metabonomics”: understanding the metabolic responses of living systems to pathophysiological stimuli via multivariate statistical analysis of biological NMR spectroscopic data.Xenobiotica. 1999; 29: 1181-1189Crossref PubMed Google Scholar) and metabolomics (a comprehensive characterization of the metabolic complement of the cell25Oliver S.G. Winson M.K. Kell D.B. et al.Systematic functional analysis of the yeast genome.Trends Biotechnol. 1998; 16: 373-378Abstract Full Text Full Text PDF PubMed Scopus (804) Google Scholar)MetabolitesIntermediary phenotypes related to metabolism in the absence of genetic informationObservation of effects from host genetics, lifestyle, and environment (including microbiome)Structural assignment can be challengingTrade-off between targeted and untargeted approaches (ie, precise knowledge of detected metabolites vs extensive coverage of the metabolome)Metabolome-wide association studies and metabolomic GWASPatient stratification through pharmacometabonomic approachesPerioperative applications Open table in a new tab How Are Metabolic Phenotypes Detected?The systematic study of metabolism and metabolic response to stimuli, or metabonomics, is usually achieved using nuclear magnetic resonance (NMR) spectroscopy and mass spectrometry (MS)23Nicholson J.K. Lindon J.C. Holmes E. “Metabonomics”: understanding the metabolic responses of living systems to pathophysiological stimuli via multivariate statistical analysis of biological NMR spectroscopic data.Xenobiotica. 1999; 29: 1181-1189Crossref PubMed Google Scholar, 24Nicholson J.K. Connelly J. Lindon J.C. et al.Metabonomics: a platform for studying drug toxicity and gene function.Nat Rev Drug Discov. 2002; 1: 153-161Crossref PubMed Scopus (1643) Google Scholar and requires statistical analysis of the spectra (Figure 1 and Table 1).NMR and MSHigh-resolution NMR spectroscopy has been used to identify chemical structures for decades and is now used extensively in metabolic studies. The robustness and reproducibility of 1H NMR spectroscopy data, combined with the capacity for widespread acquisition of reference spectra, make NMR spectroscopy a reliable technique for identification of structures in biofluids (cell media, urine, plasma, and so on), intact biopsy specimens, and tissues with minimal sample preparation. It is also possible to use 2-dimensional NMR spectroscopy, including heteronuclear 1H-13C NMR, to resolve complex 1H NMR signals along a second 13C dimension.27Dumas M.-E. Canlet C. André F. et al.Metabonomic assessment of physiological disruptions using 1H-13C HMBC-NMR spectroscopy combined with pattern recognition procedures performed on filtered variables.Anal Chem. 2002; 74: 2261-2273Crossref PubMed Scopus (0) Google Scholar The main benefit of MS over NMR is the high level of sensitivity of MS; it can detect metabolites in very low concentrations. To improve spectral resolution, MS is usually coupled to a separation step that involves gas or liquid chromatography (LC). MS/MS systems such as triple quadrupole detectors are essential for multiple reaction monitoring to ensure accurate quantification of several compounds in parallel using labeled standards to provide large-scale absolute quantification.23Nicholson J.K. Lindon J.C. Holmes E. “Metabonomics”: understanding the metabolic responses of living systems to pathophysiological stimuli via multivariate statistical analysis of biological NMR spectroscopic data.Xenobiotica. 1999; 29: 1181-1189Crossref PubMed Google Scholar, 24Nicholson J.K. Connelly J. Lindon J.C. et al.Metabonomics: a platform for studying drug toxicity and gene function.Nat Rev Drug Discov. 2002; 1: 153-161Crossref PubMed Scopus (1643) Google Scholar, 28Gieger C. Geistlinger L. Altmaier E. et al.Genetics meets metabolomics: a genome-wide association study of metabolite profiles in human serum.PLoS Genet. 2008; 4: e1000282Crossref PubMed Scopus (494) Google Scholar, 29Nicholson J.K. Holmes E. Lindon J.C. et al.The challenges of modeling mammalian biocomplexity.Nat Biotechnol. 2004; 22: 1268-1274Crossref PubMed Scopus (322) Google Scholar, 30Nicholson J.K. Holmes E. Wilson I.D. Gut microorganisms, mammalian metabolism and personalized health care.Nat Rev Microbiol. 2005; 3: 431-438Crossref PubMed Scopus (699) Google Scholar, 31Holmes E. Wilson I.D. Nicholson J.K. Metabolic phenotyping in health and disease.Cell. 2008; 134: 714-717Abstract Full Text Full Text PDF PubMed Scopus (526) Google Scholar The use of labeled standards also increases intralaboratory and interlaboratory reproducibility.32Benton H.P. Want E. Keun H.C. et al.Intra- and interlaboratory reproducibility of ultra performance liquid chromatography-time-of-flight mass spectrometry for urinary metabolic profiling.Anal Chem. 2012; 84: 2424-2432Crossref PubMed Scopus (0) Google ScholarUntargeted and Targeted Metabolic ProfilingThere are 2 approaches to metabolic profiling: hypothesis-free untargeted profiling, which presents a high potential for discovery of new biomarkers, and targeted profiling, which focuses on accurate quantification of a subset of metabolites. The use of 1H NMR for untargeted profiling can typically identify about 200 high-abundance metabolites in a particular data set, out of 1000 likely to be observed in biological samples, whereas targeted profiling can typically quantify about 80 metabolites per sample. Typically, more than 5000 uncharacterized metabolic features (corresponding to unassigned metabolites, potentially new metabolites) can be measured using ultra-performance LC coupled to quadrupole time-of-flight MS in the untargeted mode. However, most advanced targeted profiling assays claim to measure about 600 metabolites by matching standard LC-MS acquisitions with a pure compound reference database acquired under the same conditions.Using standards labeled with isotopes (such as 13C or 2H), approximately 200 metabolites can be quantified from a single sample. As a result, targeted profiling is becoming increasingly appealing to clinicians and epidemiologists, because service companies offer reliable semiquantifications for 200 to 600 metabolites detectable by MS. However, associations discovered are limited to metabolites covered by the targeted assays. In contrast, untargeted profiling provides a comprehensive analysis of the metabolome with a high potential for discovery of novel associations, even though detailed structural assignment remains time consuming. The current trend leans toward the development of comprehensive profiling strategies, which include untargeted and targeted assays.Statistical AnalysesSpectral data are very dense with thousands of signals, so mathematical modeling is necessary to highlight metabolites significantly affected by the experimental design. Many approaches have been introduced, from parallel univariate tests to multivariate statistics. One advantage of multivariate statistical analysis is that it provides information on multiple signals that can be mined to identify biomarkers of clinical use. Approaches to multivariate statistics of metabolic profiling data involve focusing on the variation within the spectral data only (ie, unsupervised) and investigating variations between the spectral data and other types of data, such as those associated with specific disease states (ie, supervised).33Fonville J.M. Richards S.E. Barton R.H. et al.The evolution of partial least squares models and related chemometric approaches in metabonomics and metabolic phenotyping.J Chemometr. 2010; 24: 636-649Crossref Scopus (0) Google Scholar Most articles use principal component analysis as an unsupervised strategy, compressing the original data set into a set of principal components made of projections of individuals (scores) and model coefficients (loadings), and partial least squares regressions as a supervised approach. Partial least squares regression presents a series of advantages for a reliable and flexible analysis of NMR and MS data.33Fonville J.M. Richards S.E. Barton R.H. et al.The evolution of partial least squares models and related chemometric approaches in metabonomics and metabolic phenotyping.J Chemometr. 2010; 24: 636-649Crossref Scopus (0) Google Scholar These statistical methods allow the identification of a set of significantly affected metabolites corresponding to the metabolic signature of the condition.Metabolic signatures derived from NMR and MS are often complex and typically contain hundreds or thousands of metabolites. The interpretation of such signatures tends to be a difficult task that can be made objective through systems biology.From Biomarkers to Metabolic PathwaysIdentification of a metabolite of interest is a good starting point, but it is not systematic and does not take advantage of the full signature. It is possible to test whether the metabolites present in a given metabolic signature belong to a particular pathway (Figure 2). Metabolite set enrichment analysis (MSEA)34Xia J. Wishart D.S. MSEA: a web-based tool to identify biologically meaningful patterns in quantitative metabolomic data.Nucleic Acids Res. 2010; 38: W71-W77Crossref PubMed Scopus (347) Google Scholar, 35Pontoizeau C. Fearnside J.F. Nayratil V. et al.Broad-ranging natural metabotype variation drives physiological plasticity in healthy control inbred rat strains.J Proteome Res. 2011; 10: 1675-1689Crossref PubMed Scopus (0) Google Scholar, 36Xia J. Wishart D.S. Web-based inference of biological patterns, functions and pathways from metabolomic data using MetaboAnalyst.Nat Protoc. 2011; 6: 743-760Crossref PubMed Scopus (626) Google Scholar, 37Kamburov A. Cavill R. Ebbels T.M.D. et al.Integrated pathway-level analysis of transcriptomics and metabolomics data with IMPaLA.Bioinformatics. 2011; 27: 2917-2918Crossref PubMed Scopus (203) Google Scholar is the metabolomic counterpart of gene set enrichment analysis38Subramanian A. Tamayo P. Mootha V.K. et al.Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles.Proc Natl Acad Sci U S A. 2005; 102: 15545-15550Crossref PubMed Scopus (19948) Google Scholar and compares the molecular content of a metabolic signature with metabolites found in known metabolic pathway maps or databases. For example, if only one tricarboxylic acid cycle intermediate is significantly increased in a sample, it is unlikely that there is a coordinated change at the pathway level. However, if 5 tricarboxylic acid cycle intermediates are significantly increased, it is more likely that the tricarboxylic acid cycle has been significantly affected and that there is a coherent up-regulation at the pathway level. However, visualization of one pathway is not sufficient, because metabolites can belong to several pathways. To generalize this approach, metabolites present in a disease signature are systematically compared with exhaustive databases of metabolites and metabolic pathways, such as the Kyoto Encyclopedia for Genes and Genomes database.39Ogata H. Goto S. Sato K. et al.KEGG: Kyoto Encyclopedia of Genes and Genomes.Nucleic Acids Res. 1999; 27: 29-34Crossref PubMed Scopus (2807) Google Scholar Statistical tests can then be performed for metabolic pathway enrichment using recently developed MSEA.34Xia J. Wishart D.S. MSEA: a web-based tool to identify biologically meaningful patterns in quantitative metabolomic data.Nucleic Acids Res. 2010; 38: W71-W77Crossref PubMed Scopus (347) Google Scholar, 35Pontoizeau C. Fearnside J.F. Nayratil V. et al.Broad-ranging natural metabotype variation drives physiological plasticity in healthy control inbred rat strains.J Proteome Res. 2011; 10: 1675-1689Crossref PubMed Scopus (0) Google Scholar, 36Xia J. Wishart D.S. Web-based inference of biological patterns, functions and pathways from metabolomic data using MetaboAnalyst.Nat Protoc. 2011; 6: 743-760Crossref PubMed Scopus (626) Google Scholar, 37Kamburov A. Cavill R. Ebbels T.M.D. et al.Integrated pathway-level analysis of transcriptomics and metabolomics data with IMPaLA.Bioinformatics. 2011; 27: 2917-2918Crossref PubMed Scopus (203) Google Scholar A statistic describes the strength of the association between the two (usually χ2 or hypergeometric analysis). In this way, a large list of potential target metabolites can be rapidly reduced into a handful of metabolic pathways, allowing for interpretation of the metabolic signature as a whole.Figure 2MSEA. Metabolites that change during development of disease form a complex metabolic pattern (A), which can be difficult to interpret. Using previous knowledge about metabolic pathways compiled in a database (B), the complex metabolic pattern is then mapped onto the metabolic pathways (C). The metabolic pattern is repeatedly tested for association with each pathway in the database (D) using enrichment tests such as χ2 analysis in the case of a qualitative overrepresentation analysis or for a quantitative enrichment analysis.35Pontoizeau C. Fearnside J.F. Nayratil V. et al.Broad-ranging natural metabotype variation drives physiological plasticity in healthy control inbred rat strains.J Proteome Res. 2011; 10: 1675-1689Crossref PubMed Scopus (0) Google Scholar, 36Xia J. Wishart D.S. Web-based inference of biological patterns, fu