Constructing a Taxonomic Model for Determining Agricultural Exposure Based on the Gut Microbiome

Abstract

Predictive models take existing information about a research area or a population of interest to predict possible outcomes. These models are widely applied in medical research and public health and are constructed using innovative machine learning tools and various modeling techniques. An individual’s gut microbial composition is affected by several environmental exposure factors, and there is a deficit in models that make predictions based on microbiome composition. In this study, we developed a method for constructing a taxonomic model that characterizes a population using previously established microbiome data as a reference. Our method employs core microbiome and indicator species analyses to find unique taxa representative of a condition of interest. The abundance and prevalence of the identified model taxa can then be measured in a new dataset to categorize the generated model based on whether it fits the condition of interest. We demonstrate our pipeline by building a taxonomic model that determines whether or not an individual has lived on a farm via model taxa, then assess its accuracy using an unrelated dataset. Our method is easily diversifiable and can be applied in medical and public health research to create taxonomic models for stratifying individuals by a characteristic of interest, such as environmental exposure, disease state and progression.