A Bayesian inference method for the analysis of transcriptional regulatory networks in metagenomic data

Abstract

BackgroundMetagenomics enables the analysis of bacterial population composition and the study of emergent population features, such as shared metabolic pathways. Recently, we have shown that metagenomics datasets can be leveraged to characterize population-wide transcriptional regulatory networks, or meta-regulons, providing insights into how bacterial populations respond collectively to specific triggers. Here we formalize a Bayesian inference framework to analyze the composition of transcriptional regulatory networks in metagenomes by determining the probability of regulation of orthologous gene sequences. We assess the performance of this approach on synthetic datasets and we validate it by analyzing the copper-homeostasis network of Firmicutes species in the human gut microbiome.ResultsAssessment on synthetic datasets shows that our method provides a robust and interpretable metric for assessing putative regulation by a transcription factor on sets of promoter sequences mapping to an orthologous gene cluster. The inference framework integrates the regulatory contribution of secondary sites and can discern false positives arising from multiple instances of a clonal sequence. Posterior probabilities for orthologous gene clusters decline sharply when less than 20 % of mapped promoters have binding sites, but we introduce a sensitivity adjustment procedure to speed up computation that enhances regulation assessment in heterogeneous ortholog clusters. Analysis of the copper-homeostasis regulon governed by CsoR in the human gut microbiome Firmicutes reveals that CsoR controls itself and copper-translocating P-type ATPases, but not CopZ-type copper chaperones. Our analysis also indicates that CsoR frequently targets promoters with dual CsoR-binding sites, suggesting that it exploits higher-order binding conformations to fine-tune its activity.ConclusionsWe introduce and validate a method for the analysis of transcriptional regulatory networks from metagenomic data that enables inference of meta-regulons in a systematic and interpretable way. Validation of this method on the CsoR meta-regulon of gut microbiome Firmicutes illustrates the usefulness of the approach, revealing novel properties of the copper-homeostasis network in poorly characterized bacterial species and putting forward evidence of new mechanisms of DNA binding for this transcriptional regulator. Our approach will enable the comparative analysis of regulatory networks across metagenomes, yielding novel insights into the evolution of transcriptional regulatory networks.Electronic supplementary materialThe online version of this article (doi:10.1186/s13015-016-0082-8) contains supplementary material, which is available to authorized users.

DRAM for distilling microbial metabolism to automate the curation of microbiome function

“…Consequently, DRAM is only a first step in identifying key functional genes, as subsequent non-homology based methods (e.g. phylogenetic analyses, protein modeling ( 73 ), gene synteny, Bayesian inference framework ( 74 , 75 )) or physiological or biochemical characterization are often required to validate findings from any homology-based annotator.…”

Flexible comparative genomics of prokaryotic transcriptional regulatory networks

“…However, this approach is not well-suited for the comparative genomics framework, because thresholds may often need to be tuned in different bacterial genomes owing to their particular distribution of oligomers [6]. To circumvent this problem, here we adopt a Bayesian probabilistic framework originally developed for regulon analysis in metagenomic sequences [26]. This framework estimates posterior probabilities of regulation that are easily interpretable and directly comparable across species.…”

“…For each position i of a promoter region, we first combine the PSSM scores obtained in the forward (f) and reverse (r) strands using the function [26]:…”

Why Are Omics Technologies Important to Understanding the Role of Nutrition in Inflammatory Bowel Diseases?

“…The colonic microbiota plays an even more critical role in IBD pathogenesis. Metagenomics is the current technology which describes the microbial genomic composition of the GI tract, usually measured from a faecal sample [61]. Changes in the composition of these bacterial populations, and/or of their collective genome, can have important implications for disease susceptibility, including intestinal inflammation [62,63].…”

“…In addition, RNA-seq approaches allow a measurement of the expression of microbial genes to be measured. Hobbs et al, [61] developed a statistical method to characterise transcriptional regulatory networks from metagenomics data. Such approaches allow an assessment of the complex interplay between host and microbial genes, and their impact on the health of the host.…”