Strain-specific gene function prediction and the Human Microbiome Bioactives Resource

Most genetic material encountered in microbial communities - including the human microbiome - remains both uncharacterized and, often, strain-specific. This leaves a tremendous gap in translating microbiome science to personalized biomarkers and therapies. While multi-omic studies represent a promising means of prioritizing leads for validation, this still leaves thousands of molecular targets for most phenotypes of interest. The Human Microbiome Bioactives Resource (HMBR) provides a set of computational and experimental protocols for accelerating these tasks, including the identification of “bioactive” strains, microbial proteins, and chemical products from the human gut, as well as newly predicting function for uncharacterized genes and metabolites. These have proven effective in applications as diverse as the inflammatory bowel diseases (IBD, in a meta-analysis spanning over 8,700 metagenomes) or companion animal nutrition (employing almost 2,500 canine and feline gut samples). In an end-to-end example linking metagenomics, metatranscriptomics, metabolomics, inflammation phenotypes, and drug treatment, we have specifically identified a subset of microbial gene families responsible for 5-ASA inactivation, the leading cause of progression to second-line treatments in IBD. Together, this family of approaches provides multiple new avenues for identifying molecular activities responsible for microbiome-phenotype associations.