📝 SummarXiv

Abstract

The gut microbiome significantly influences responses to cancer therapies, including immunotherapies, primarily through its impact on the metabolome. Despite some studies on effects of specific microbial genera on individual metabolites, there is little prior work identifying key microbiome components at the genus level that shape the overall metabolome profile. To address this gap, we introduce B-MASTER (Bayesian Multivariate regression Analysis for Selecting Targeted Essential Regressors), a fully Bayesian framework with an L1 penalty to promote sparsity and an L2 penalty to shrink coefficients for non-major covariates, thereby isolating essential regressors. The method is paired with a scalable Gibbs sampling algorithm, whose computation grows linearly with the number of parameters and remains largely unaffected by sample size for models of fixed dimensions. Notably, B-MASTER enables full posterior inference for models with up to four million parameters within a practical time-frame. Its theoretical guarantees include posterior contraction, selection consistency, and robustness under mild misspecification. Using this approach, we identify key microbial genera shaping the metabolite profile, analyze their effects on the most abundant metabolites, and investigate metabolites differentially abundant in colorectal cancer (CRC) patients. These results provide foundational insights into microbiome-metabolite relationships relevant to cancer, a connection largely unexplored in existing literature.