A probabilistic model to identify the core microbial community

Abstract

The core microbial community has been hypothesized to have essential functions ranging from maintaining health in animals to protection against plant disease. However, the identification of the core microbial community is frequently based on arbitrary thresholds, selecting only the most abundant microorganisms. Here, we developed and tested an approach to identify the core community based on a probabilistic model. The Poisson distribution was used to identify OTUs with a probable occurrence in every sample of a given dataset. We identified the core communities of four extensive microbial datasets, and compared the results with conventional, but arbitrary, methods. The datasets were composed of the microbiomes of humans (tongue, gut, and skin), mice (gut), plant (grapevine) tissue, and the maize rhizosphere. Our proposed method revealed core microbial communities with higher richness and diversity than those previously described. This method also includes a greater number of rare taxa in the core, which are often neglected by arbitrary threshold methods. We demonstrated that our proposed method revels a probable core microbial community for each different habitat, which extend our knowledge about shared microbial communities. Our proposed method may help the next steps proving the essential functions of core microbial communities.\n\nOriginality-Significance StatementMore rigorous and less arbitrary statistical methods could increase knowledge regarding the role of microorganisms and their interactions. Here, we suggest a probabilistic method to identify the microbial core community across systems. Our method identifies a large proportion of the rare community that likely belongs to the microbial core community, which was not identified by conventional methods. Our probabilistic model is a non-arbitrary approach to defining the microbial core community, which may help in the next step of the microbial core community studies.