Meta-analysis of targeted metabolomics data from heterogeneous biological samples provides insights into metabolite dynamics
IntroductionMass spectrometry-based metabolomics coupled to liquid chromatography, or LC-MS metabolomics, has become the most popular tool for global metabolite abundance profiling to study metabolism. However, the physicochemical complexity of metabolites poses a major challenge for reliable measurements of metabolite abundance. One way to address the issue is to use multiple chromatographic methods to capture a greater range of molecular diversity. We previously developed a tandem mass spectrometry-based label-free targeted metabolomics analysis framework coupled to two distinct chromatographic methods, reversed-phase liquid chromatography (RPLC) and hydrophilic interaction liquid chromatography (HILIC), with dynamic multiple reaction monitoring (dMRM) for simultaneous detection of over 200 metabolites to study core metabolic pathways. ObjectivesWe aim to analyze a large-scale heterogeneous data compendium generated from our LC-MS/MS platform with both RPLC and HILIC methods to systematically assess measurement quality in biological replicate groups and to gain insights into metabolite dynamics across different biological conditions. MethodsOur metabolomics framework was applied in a wide range of experimental systems including cancer cell lines, tumors, extracellular media, primary cells, immune cells, organoids, organs (e.g. pancreata), tissues, and sera from human and mice. We also developed computational and statistical analysis pipelines, which include hierarchical clustering, replicate-group CV analysis, correlation analysis, and case-control paired analysis. ResultsWe generated a compendium of 42 heterogeneous deidentified datasets with 635 samples using both RPLC and HILIC methods. There exist signature metabolites that correspond to heterogeneous phenotypes, involved in several metabolic pathways. The RPLC method shows overall better reproducibility than the HILIC method for most metabolites including polar amino acids. Correlation analysis reveals high confidence metabolites irrespective of experimental systems such as methionine, phenylalanine, and taurine. We also identify homocystine, reduced glutathione, and phosphoenolpyruvic acid as highly dynamic metabolites across all case-control paired samples. ConclusionsOur study is expected to serve as a resource and a reference point for a systematic analysis of label-free LC-MS/MS targeted metabolomics data in both RPLC and HILIC methods with dMRM.