Researchers from McMaster University and the Massachusetts Institute of Technology (MIT) made two scientific breakthroughs at once. Not only did they discover the latest antibiotics targeting inflammatory bowel disease (IBD), they also managed to use a new type of AI to accurately predict how drugs work. To their knowledge, this is the global first of AI.
Details in the Journal on October 3, 2025 Natural MicrobiologyThis discovery uncovers promising new treatment options for millions of people affected by Crohn's disease and other related conditions, and introduces important new applications of AI in drug discovery research.
“This work shows that, as far as AI-inducing agents are discovered, it still hurts the surface,” said Jon Stokes, assistant professor in the Department of Biochemistry and Biomedical Sciences at McMaster and lead investigator of the new study. “The development of new drugs designed to target IBD has been quickly tracked thanks to the collaboration between humans and generator AI.”
Antibiotics for IBD
Most of the antibiotics used in clinics today are “broad spectrum” drugs. This means that it wipes out the good bacteria as well as the bacteria that cause disease. “It's the core,” explains Stokes.
This could create opportunities for invasive, drug-resistant bacteria. E. coli, to migrate and colonize the intestines that can exacerbate clone-like conditions.
However, enterolin, a new antibiotic discovered in McMaster, is a “narrow spectrum” drug that attacks the microbiota with spares, attacking only certain groups of bugs that cause disease. Intestinal bacteria, which happens to contain E. coli.
This means it's not just killing E. coliHowever, drug-resistant strains reduce the chances of colonizing the intestines in the first place.
“This new drug is a very promising treatment candidate for the millions of patients living with IBD,” says Stokes, a teacher. “We don't currently have treatments for these conditions, so developing something that could mean symptoms will help people experience a higher quality of life.”
How do drugs work? Ask AI
Until now, AI has been used as a tool to predict which molecules may have therapeutic potential, but in this study, it was used to explain what researchers call “mechanism of action” (MOA) or how to call it. Drug attack disease.
“AI has accelerated the speed at which new drug candidates can explore the chemical space, but up until now, it has rarely alleviated the major bottlenecks of drug development that these new drug candidates understand what they are doing,” explains Stokes.
According to him, MOA research is essential for drug development. They help scientists to check safety, optimize doses, make modifications to improve efficacy, and uncover all-new drug targets. It also helps regulators determine whether a particular drug candidate is suitable for human use.
But they are also highly famous – and slow.
Stokes says that thorough MOA research takes up to two years and costs around $2 million. However, using AI, his group did Enterolin in just six months, for just $60,000.
Indeed, after his lab discovered a new antibiotic, Stokes, who connected with colleagues at MIT's Computer Science and Artificial Intelligence Lab (CSAIL), saw whether any of their emerging machine learning platforms would help him quickly track his upcoming MOA research.
In just 100 seconds, he was given a prediction. His new drug attacked a microscopic protein complex called LOLCDE, which is essential for the survival of certain bacteria.
“We've got a lot of effort into making predictions,” said Regina Barzilay, MIT's engineering professor and developer of AI models. “What we're showing here is that AI can provide a mechanical explanation, which is important for moving molecules through the development pipeline.”
Barzilay was recently listed Time MagazineThe most influential people in AI.
Stokes emphasizes that while predictions are intriguing, it is exactly that. He still has to do traditional MOA research in the lab.
“We cannot currently assume that these AI models are completely correct, but that's what I did it Remove the speculation from the next step.” Michael G. DeGroote Institute for Fectious Disease Research With McMaster.
His team then began investigating Enterolin's MOA, led primarily by McMaster's graduate student Dennis Catactan, using MIT's predictions as a starting point.
Within just a few months it was clear that AI was actually right.
“We did all the standard MOA work-inspections to verify our predictions. We checked whether the experiment would back up AI. “By doing this, we've cut our normal timeline for a year and a half.”
This has led to Stokes successfully using AI Discover viable drug candidates, Fast efforts in global drug discoveryand decide how the new drug works. But if you ask him, he will tell you it to you – it's profitable as it is – AI is just a means of end.
“Drug resistance and our lack of new drugs are leaky faucets,” he says. “You can leave it for a while, but in the end there's a big problem. AI is my wrench. It's a tool to fix leaks before it gets flooded. That's really it. My only focus is taking new medications to patients who need them.
Route to the patient
Stokes spin-out company, Painful Biohas already approved enterolin from McMaster and is currently optimised for human use.
The company is also testing modified versions of new antibiotics against other drug-resistant bacteria. Klebsiellaand the early results are promising.
“The identification of enterolin highlights the notable science that appears in McMaster,” said Jeff Skinner, CEO of Stoked Bio. “We are proud to partner with the university by translating this breakthrough into actual treatments for patients.”
If everything goes well, Stokes says the new drug will be ready for human testing within three years. His research team wants to meet.
“It's surreal to work on translations like this,” says Kataktan. “The fact that what we discover in our lab might one day help patients is truly amazing and really reinforces the importance of our work.”
reference: Catacutan DB, Tran V, Arnold A, et al. Discovery of narrow spectral antibiotics and elucidation of the mechanisms of artificial intelligence induction. Nat Microbiol. 2025. doi:10.1038/s41564-025-02142-0
This article has been republished from: Note: Materials may have been edited for length and content. For more information, please contact the source quoted. You can access the press release publishing policy here.
