THE HUMAN MICROBIOME DRUG METABOLISM DATABASE

Abstract

We rely on oral drugs to treat several diseases and infections. Yet, the gut microbiome modifies oral drugs within the human gut by using enzymes, facilitating efficient chemical reactions. These drug modifications impact effective doses and outcomes for individuals. The gut microbiome can convert drugs destined for excretion back to active drugs, and the converse is also true, the microbiome can inactivate active drugs, and both may lead to toxic effects. There is no resource for cataloging bacterial drug-metabolizing genes within the human gut microbiome with analytical tools to annotate these genes in sequenced gut microbiomes. I created a resource called the Human Microbiome Drug Metabolism (HMDM) database. I analyzed 1,196 unpublished sequenced gut bacterial genomes from 8 healthy adult donors to predict genes that encode enzymes capable of metabolizing drugs using two in silico methods I developed, namely MAGIS and AutoPhylo. I reviewed the scientific literature and built an ontology-centric database, the HMDM, to catalog the bacterial drug-metabolizing genes and drugs they modify. I developed DME software to predict bacterial genes capable of metabolizing host-directed drugs using the HMDM data. We experimentally validated four novel AMR gene homologs predicted from the genomes. The HMDM is curated with 50 genes reported to metabolize drugs and 45 gene variants of the β-glucuronidase (uidA) gene. MAGIS was used to predict 246 putative bacterial drug-metabolizing genes. I predicted the three novel AMR gene homologs that resemble fosfomycin thiol transferase enzymes using AutoPhylo. The MIC experiment shows that fosD1, fosD2, and fosD3 have MIC of 8μg/mL, 8μg/mL, and >512μg/mL, respectively. The genes fosD1 and fosD2 are of unknown function, and FosD3 converts fosfomycin. The HMDM database is limited to bacterial genes. The in silico methods are critical for studying bacterial drug metabolism to predict drug fate and patient outcomes.