Researchers from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) and McMaster University have discovered a new antibiotic compound called enterololin that selectively targets harmful gut bacteria linked to Crohn’s disease flare-ups, while preserving the rest of the beneficial microbiome. This precision approach could offer a safer alternative to broad-spectrum antibiotics, which often disrupt the delicate balance of gut microbes and may worsen inflammatory bowel disease over time. Using a generative AI model called DiffDock, developed by MIT PhD student Gabriele Corso and Professor Regina Barzilay, the team rapidly identified the molecular mechanism of action for enterololin—something that traditionally takes years. DiffDock treats molecular docking as a probabilistic problem, using a diffusion model to iteratively refine predictions of how a small molecule binds to a protein. Within minutes, the model predicted that enterololin binds to a protein complex called LolCDE, which is essential for transporting lipoproteins in certain bacteria. The prediction was validated through multiple lab experiments. The researchers created enterololin-resistant strains of E. coli, which showed mutations in the lolCDE gene exactly where the AI had predicted. RNA sequencing revealed disruptions in lipoprotein transport pathways, and CRISPR gene editing confirmed that knocking down lolCDE made bacteria more sensitive to the drug. These findings aligned perfectly with the AI’s output, confirming the mechanism with high confidence. This breakthrough cut the typical timeline for mechanism-of-action studies—from 18 months to two years down to just six months—while significantly reducing costs. The approach demonstrates how AI can go beyond identifying promising molecules and instead provide actionable biological insights that accelerate drug development. In mouse models of Crohn’s-like inflammation, enterololin reduced levels of disease-causing E. coli and helped animals recover faster, with better microbiome stability compared to treatment with vancomycin. The results suggest a promising path toward targeted therapies that treat infections without harming the broader microbial community. The compound is now being developed by Stoked Bio, a spinout company led by Jon Stokes, the senior author of the study. The team is optimizing enterololin’s properties for human use and exploring its potential against other drug-resistant pathogens like Klebsiella pneumoniae. Clinical trials could begin within a few years. Experts outside the study, including Yves Brun from the University of Montreal, praised the work as a powerful example of AI’s growing role in combating antimicrobial resistance. The research was supported by multiple organizations, including the Weston Family Foundation, Canadian Institutes of Health Research, the Jameel Clinic, and the U.S. Defense Threat Reduction Agency. The team has made their sequencing data publicly available and released the DiffDock-L code on GitHub, enabling wider access to this transformative tool. The study marks a significant step toward a new era of precision antibiotics, where AI helps uncover not just what a drug does, but how and why it works—unlocking faster, smarter, and more targeted treatments for a range of diseases.