The compound, known as enterololin, represents a promising new class of precision antibiotics. Unlike traditional broad-spectrum drugs that wipe out both harmful and beneficial gut bacteria, enterololin zeroes in on a specific bacterial culprit — a strain of E. coli linked to Crohn’s flare-ups — while leaving the rest of the gut ecosystem largely untouched.
In tests on mice with Crohn’s-like inflammation, enterololin not only targeted the harmful bacteria but also helped the animals recover faster and maintain a healthier microbiome compared to those treated with vancomycin, a widely used antibiotic.
The discovery was made possible through the use of artificial intelligence, specifically a generative AI tool called DiffDock, developed at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL). The model was instrumental in identifying how enterololin interacts with bacterial proteins — a process that typically takes years of laboratory work, but which researchers accomplished in just months.
“The problem isn’t finding molecules that kill bacteria,” explained Jon Stokes, the study’s senior author and assistant professor at McMaster University. “It’s understanding what those molecules actually do inside bacteria — that’s what allows you to turn them into safe, effective therapies.”
DiffDock, designed by MIT researchers Gabriele Corso and Regina Barzilay, predicts how small molecules bind to protein targets inside bacteria. In this study, the AI model rapidly pinpointed enterololin’s target — a bacterial protein complex called LolCDE, responsible for transporting essential lipoproteins. Laboratory experiments later confirmed the AI’s prediction through a series of genetic, RNA sequencing, and CRISPR-based tests.
“When computational predictions and wet-lab data converge on the same mechanism, that’s when you know you’re onto something real,” Stokes noted.
Barzilay, who co-leads MIT’s Jameel Clinic for Machine Learning in Health, said the project illustrates a broader shift in how AI is transforming life sciences research.
Caption:By using AI to sift through more than 10,000 molecules, researchers found enterololin (inset), a compound that blocks a key pathway in harmful gut bacteria and, in mice with IBD, eased infection without disturbing the rest of the microbiome. Image credit: Alex Shipps/MIT CSAIL, using assets from the researchers and Pexels
“AI is no longer just about discovering new molecules,” she said. “It’s now helping us understand how they work, which is the critical step in turning a promising molecule into a viable medicine.”
By integrating AI-driven modeling with laboratory science, the MIT–McMaster team shortened the typical drug discovery timeline from up to two years to roughly six months — and at a fraction of the cost.
Although enterololin is still in early development, progress is already being made toward clinical translation. Stokes’ startup, Stoked Bio, has licensed the compound and is refining its properties for potential human trials. Researchers are also exploring variations of the molecule for use against other resistant bacteria, such as Klebsiella pneumoniae.
Experts say the implications could be far-reaching. Narrow-spectrum antibiotics have long been a goal in medicine — treatments that can fight infection without disrupting the body’s natural microbiome — but they’ve been notoriously difficult to develop. AI tools like DiffDock could finally make that goal achievable, ushering in a new era of targeted antimicrobial therapy.
For Crohn’s disease patients, the potential benefits are especially significant: a treatment that eases inflammation without worsening gut imbalance. On a broader scale, precision antibiotics like enterololin could play a crucial role in combating the global rise of antimicrobial resistance.
“This study uses a powerful and elegant combination of AI methods to uncover how a new antibiotic works,” said Yves Brun, a microbiology professor at the University of Montreal who was not involved in the research. “It’s a glimpse into how AI can become a vital ally in our fight against resistant bacteria.”
If further studies succeed, enterololin could mark the beginning of a new chapter in antibiotic development — one where data-driven precision replaces broad destruction, and where artificial intelligence helps medicine outthink one of humanity’s oldest biological foes.