Microbiota-derived metabolites significantly impact host biology, prompting extensive research on metabolic shifts linked to the microbiota. Recent studies have explored both direct metabolite analyses and computational tools for inferring metabolic functions from microbial shotgun metagenome data. However, no existing tool specifically focuses on predicting changes in individual metabolite levels, as opposed to metabolic pathway activities, based on shotgun metagenome data. Understanding these changes is crucial for directly estimating the metabolic potential associated with microbial genomic content. We introduce PredCMB (Predicting Changes in Microbial metaBolites), a novel method designed to predict alterations in individual metabolites between conditions using shotgun metagenome data and enzymatic gene-metabolite networks. PredCMB evaluates differential enzymatic gene abundance between conditions and estimates its influence on metabolite changes. To validate this approach, we applied it to two publicly available datasets comprising paired shotgun metagenomics and metabolomics data from inflammatory bowel disease (IBD) cohorts and the cohort of gastrectomy for gastric cancer. Benchmark evaluations revealed that PredCMB outperformed a previous method by demonstrating higher correlations between predicted metabolite changes and experimentally measured changes. Notably, it identified metabolite classes exhibiting major alterations between conditions. By enabling the prediction of metabolite changes directly from shotgun metagenome data, PredCMB provides deeper insights into microbial metabolic dynamics than existing methods focused on pathway activity evaluation. Its potential applications include refining target metabolite selection in microbial metabolomic studies and assessing the contributions of microbial metabolites to disease pathogenesis. Freely available to non-commercial users at https://www.sysbiolab.org/predcmb. Supplementary data are available at Bioinformatics online.
Read full abstract