B-MASTER: Scalable Bayesian Multivariate Regression for Master Predictor Discovery in Colorectal Cancer...

Area of Science:

• Microbiome research
• Metabolomics
• Cancer therapy

Background:

• The gut microbiome significantly impacts cancer treatment efficacy by modulating host metabolism.
• Current methods struggle to identify microbial genera that systematically control the entire metabolome.
• There's a need for scalable statistical tools to find system-level microbial regulators in high-dimensional microbiome-metabolome data.

Purpose of the Study:

• To introduce B-MASTER, a novel Bayesian multivariate regression framework.
• To identify microbial genera that act as master regulators of the host metabolome.
• To enable system-level analysis of microbiome-metabolome interactions.

Main Methods:

• Developed B-MASTER, a scalable Bayesian multivariate regression framework.
• Incorporated ℓ1 sparsity and ℓ2 group shrinkage for identifying cross-metabolite regulators.
• Utilized a Gibbs sampler for near-linear computational scaling, supporting large-scale models.
• Provided theoretical guarantees including posterior contraction and selection consistency.

Main Results:

• B-MASTER successfully identifies key microbial genera regulating global and cancer-associated metabolite patterns.
• The framework reveals system-level regulatory structures within the microbiome-metabolome axis.
• Analysis of colorectal cancer data demonstrated the method's capability in uncovering significant microbial drivers.

Conclusions:

• B-MASTER offers a scalable solution for identifying microbial regulators in complex biological systems.
• The findings highlight the importance of system-level microbiome-metabolome interactions in cancer therapy.
• The developed framework advances our ability to understand host-microbiome-metabolome interplay.