Phylogenetically informed Bayesian truncated copula graphical models for microbial association networks

TL;DR

This paper introduces a phylogenetically informed Bayesian truncated copula graphical model for microbial association networks, effectively handling zero-inflated microbiome data and incorporating evolutionary history to improve network estimation.

Contribution

It presents a novel Bayesian model that integrates phylogenetic information and zero-inflation handling, advancing microbial network analysis methods.

Findings

Evolutionary information improves network estimation accuracy.

Identified three distinct microbial communities in gut microbiome data.

Communities differ based on oxygen utilization by microorganisms.

Abstract

Microorganisms play a critical role in host health. The advancement of high-throughput sequencing technology provides opportunities for a deeper understanding of microbial interactions. However, due to the limitations of 16S ribosomal RNA sequencing, microbiome data are zero-inflated, and a quantitative comparison of microbial abundances cannot be made across subjects. By leveraging a recent microbiome profiling technique that quantifies 16S ribosomal RNA microbial counts, we propose a novel Bayesian graphical model that incorporates microorganisms' evolutionary history through a phylogenetic tree prior and explicitly accounts for zero-inflation using the truncated Gaussian copula. Our simulation study reveals that the evolutionary information substantially improves the network estimation accuracy. We apply the proposed model to the quantitative gut microbiome data of 106 healthy subjects, and identify three distinct microbial communities that are not determined by existing microbial network estimation models. We further find that these communities are discriminated based on microorganisms' ability to utilize oxygen as an energy source.