Network-based integration and interpretation of large-scale public metabolomics datasets using iDMET and iDMET+

An increasing number of metabolomics datasets have become publicly available in repositories such as MetaboLights and Metabolomics Workbench. Integrating these datasets across studies may provide biological insights that are difficult to obtain from individual studies alone. One major challenge in integrating metabolomics datasets across studies is that peak areas are often not directly comparable between studies. In addition, different analytical platforms frequently target different subsets of metabolites, resulting in limited overlap among detected metabolites. Even if only absolute quantitative measurements are used, only a small number of metabolites are commonly measured across studies. To address these challenges, we developed iDMET, a network-based approach for integrating metabolomics datasets. Each dataset is represented as a differential metabolomic profile consisting of metabolites that are increased or decreased between two groups. Rather than directly comparing peak areas across studies, iDMET identifies similarities between differential metabolomic profiles and constructs an integrated network based on pairwise comparisons of differential metabolomic profiles. We also developed iDMET+, an enrichment analysis framework for the biological interpretation of the resulting network. To evaluate the utility of iDMET, we applied it to 27 publicly available cancer metabolomics datasets. The resulting network revealed known cancer-associated metabolic alterations across multiple studies and identified relationships between datasets that were not apparent from individual studies alone. Furthermore, iDMET+ identified biologically meaningful metabolite sets and generated hypotheses that were not readily obtained using conventional pathway analysis. These results demonstrate that iDMET and iDMET+ provide a practical framework for integrating and interpreting large-scale public metabolomics datasets.