MDiNE: A model to estimate differential co-occurrence networks in microbiome studies

Abstract

Motivation:The human microbiota is the collection of microorganisms colonizing the human body, and plays an integral part in human health. A growing trend in microbiome analysis is to construct a network to estimate the co-occurrence patterns among taxa though precision matrices. Existing methods do not facilitate investigation into how these networks change with respect to covariates. Results:We propose a new model called Microbiome Differential Network Estimation (MDiNE) to estimate network changes with respect to a binary covariate. The counts of individual taxa in the samples are modelled through a multinomial distribution whose probabilities depend on a latent Gaussian random variable. A sparse precision matrix over all the latent terms determines the co-occurrence network among taxa.

Model-based and phylogenetically adjusted quantification of metabolic interaction between microbial species

“…Mandakovic and colleagues compared microbial co-occurrence networks representing bacterial soil communities from different environments to determine the impact of a shift in environmental variables on the community’s taxonomic composition and their relationships [ 22 ]. MDiNE is another recently developed model for estimating differential co-occurrence networks in microbiome studies [ 23 ]. Notably, Faust et al [ 24 ] applied generalized boosted linear models to infer thousands of significant co-occurrence and co-exclusion relationships between 197 clades occurring throughout the human microbiomes; their study revealed reverse correlation between functional similarity and phylogenetic distance among bacterial species, which is unsurprising.…”

A network-based approach to deciphering a dynamic microbiome’s response to a subtle perturbation

“…Subtle perturbation and co-occurrence patterns. Microbes with co-occurrence patterns were known to be driven by metabolic interactions and competition for resources 26 , and had the potential to capture crucial community characteristics that might not be discovered in microbial diversity or abundance-based analysis 27,28 . Based on 25 highly correlated microbial families ( r C i B i ) with peculiar co-occurrence (ascent, descent, and convex) abundance patterns ( Fig.…”

Reply

“…Several approaches have been proposed to address co-occurrence networks including the Spearman correlation analysis used by Gupta et al 4,5 and suggested by ours colleagues Fu et al 1 These methods usually deal with a number of taxa (OTUs) large relative to the sample size and assume sparsity (an abundance of 0 entries) to achieve stable precision matrix estimation. 7 However, inadequate (small) sample sizes remain a major drawback, resulting in unstable estimates of precision matrices for which applying a penalization scheme is recommended as we and other authors previously reported. 7 Even when using Lasso as an efficient solution in the domain of sparse precision matrix estimation, having a limited sample size (<30 samples) tends to produce unstable co-occurrence data and networks to our knowledge.…”

“…7 However, inadequate (small) sample sizes remain a major drawback, resulting in unstable estimates of precision matrices for which applying a penalization scheme is recommended as we and other authors previously reported. 7 Even when using Lasso as an efficient solution in the domain of sparse precision matrix estimation, having a limited sample size (<30 samples) tends to produce unstable co-occurrence data and networks to our knowledge.…”