Expanding the drug discovery space with predicted metabolite-target interactions.
Journal:
Communications biology
Published Date:
Mar 5, 2021
Abstract
Metabolites produced in the human gut are known modulators of host immunity. However, large-scale identification of metabolite-host receptor interactions remains a daunting challenge. Here, we employed computational approaches to identify 983 potential metabolite-target interactions using the Inflammatory Bowel Disease (IBD) cohort dataset of the Human Microbiome Project 2 (HMP2). Using a consensus of multiple machine learning methods, we ranked metabolites based on importance to IBD, followed by virtual ligand-based screening to identify possible human targets and adding evidence from compound assay, differential gene expression, pathway enrichment, and genome-wide association studies. We confirmed known metabolite-target pairs such as nicotinic acid-GPR109a or linoleoyl ethanolamide-GPR119 and inferred interactions of interest including oleanolic acid-GABRG2 and alpha-CEHC-THRB. Eleven metabolites were tested for bioactivity in vitro using human primary cell-types. By expanding the universe of possible microbial metabolite-host protein interactions, we provide multiple drug targets for potential immune-therapies.
Authors
Keywords
Anti-Inflammatory Agents
Bacteria
Cells, Cultured
Data Mining
Databases, Factual
Drug Discovery
Gastrointestinal Agents
Gastrointestinal Microbiome
Gene Expression Profiling
Host-Pathogen Interactions
Humans
Inflammatory Bowel Diseases
Ligands
Machine Learning
Metabolome
Metabolomics
Molecular Targeted Therapy
Protein Interaction Maps
Signal Transduction
Transcriptome