AI predicts antibiotic's IBD mechanism–scientists later confirm (healthsci.mcmaster.ca)

Researchers at McMaster University and MIT announced the discovery of enterololin, a new narrow‑spectrum antibiotic that selectively targets Enterobacteriaceae (including E. coli) and spares the broader gut microbiome—making it a promising candidate for treating inflammatory bowel diseases (IBD) such as Crohn’s. The drug, described in Nature Microbiology (Oct 3, 2025), reduces opportunities for opportunistic, drug‑resistant strains to colonize the gut and is already licensed to Stokes’ spin‑out, Stoked Bio, which aims for human trials within three years and is testing derivatives against pathogens like Klebsiella. The team also used an AI model from MIT’s CSAIL (DiffDock, developed by Regina Barzilay) to predict enterololin’s mechanism of action (MOA) in roughly 100 seconds: the drug inhibits the LolCDE membrane protein complex essential for certain bacteria. That prediction guided traditional MOA experiments and was experimentally validated within months—cutting a process that can take ~2 years and $2M down to six months and ~$60k. This is a notable first: AI provided a mechanistic explanation, not just candidate molecules, accelerating lead validation, informing safety/optimization steps, and demonstrating how generative ML can materially compress drug‑discovery timelines while remaining a hypothesis‑generation tool requiring empirical confirmation.