AI model designs new antibiotic in early tests, accelerating drug discovery against resistant bacteria

AI model rapidly designs new antibiotic to fight resistant bacteria

McMaster UniversityApr 23 2026 Researchers at McMaster University have developed a new generative artificial intelligence (AI) model capable of drastically speeding up drug discovery - and, in early tests, it has already designed a brand-new antibiotic. The discovery is a demonstration of how AI could dramatically improve the slow and costly search for new antimicrobial medicines, as bacteria and other microbes continue to evolve resistance to our current suite of drugs. The new model, called SyntheMol-RL, is trained to explore a vast chemical space of up to 46 billion possible compounds - far beyond what could realistically be tested in the lab, where even large-scale screens top out at around a million molecules. Drawing on roughly 150,000 molecular 'building blocks' and a set of 50 chemical synthesis reactions, the AI model is designed to generate structurally novel antibiotic candidates. "In the lab, we can build chemical compounds using a set of smaller chemical fragments, which can be stuck together like molecular Lego blocks," says Assistant Professor Jon Stokes, whose laboratory developed the new model. "SyntheMol-RL configures those fragments in different ways, faster than humans ever could, to create new, larger chemical compounds that should - based on its knowledge - be antibacterial." Stokes, a member of the Michael G. DeGroote Institute for Infectious Disease Research, says that while generative AI is becoming increasingly effective at designing novel antibiotic candidates, key properties that determine a potential drug's clinical viability remain difficult to assess without extensive - and expensive - laboratory testing. "It doesn't matter if you find a new chemical that's antibacterial in the lab if it can't dissolve inside the body, if it's toxic to human cells, or if it can't be metabolized and expelled after it has done its job," he explains. "Bleach is antibacterial - so is fire. But they obviously don't tick those other boxes. Good drug candidates must meet several different criteria, otherwise they'll never become actual medicine." Past iterations of SyntheMol exclusively designed molecules with antibacterial activity, without consideration for these other critical properties. But, over the past two years, Stokes' team - with collaborators at Stanford University - has refined the model so it only generates antibacterial compounds that are easy to develop in the lab and likely to be soluble in the body. "There is a lot of conflict between compounds that are antibacterial and compounds that are water soluble," says Gary Liu, a graduate student in Stokes' lab and lead developer of the new model. "In previous studies, filtering for compounds that were both antibacterial and soluble after our prompt often left us with significantly fewer viable drug candidates, so we built solubility right into the generation process and now the model can efficiently design antibiotic candidates with greater clinical promise." In a new study, published April 23 and selected for the cover of the June issue of Molecular Systems Biology, Stokes' team put their enhanced model to the test. They tasked it with generating water-soluble antibiotics that could treat infections caused by Staphylococcus aureus - colloquially known as "staph infections" - and quickly got several hits. From a batch of 79 model-proposed antibacterials, Stokes' group homed in on one particularly interesting compound - a novel, water-soluble compound that seemed likely to have antibiotic activity against S. aureus. The new computer-designed drug candidate, which they called synthecin, was then formulated as a topical cream in the lab and tested on an otherwise drug-resistant wound infection in mouse models. "Synthecin was highly effective at controlling the infection," says Denise Catacutan, a grad student in Stokes' lab who led the wet lab portions of the study. "It worked extremely well as a topical drug, and also shows early promise as something that could be applied or optimized for systemic use in the future." While the new study highlights synthecin's promise, the team has yet to uncover how the drug inhibits bacteria - a key step, Stokes says, in determining its safety profile and therefore its likelihood of someday landing in clinics. His group is now actively engaged in these critical "mechanism-of-action" studies. Regardless of how those studies play out, though, the group sees the discovery of synthecin as validation that their AI model can rapidly generate high-potential drug candidates, shifting the burden of drug discovery from finding viable compounds to designing and optimizing them. That shift, Stokes says, is significant not only for antibiotic discovery, but also for all areas of biochemistry. "We used our model to design new antibiotics, but it's capable of so much more," says Stokes, a faculty member at the Marnix E. Heersink School of Biomedical Innovation and Entrepreneurship and an executive member of NexusHealth. "We built it to be disease agnostic, meaning it could just as easily generate novel drug candidates for diabetes or cancer or other indications." Stokes' lab continues to enhance SyntheMol and anticipates that an even more robust version will be available later this year. McMaster University Journal reference: DOI: 10.1038/s44320-026-00206-9

McMaster-Built AI Speeds Up Drug Discovery, Designs New Antibiotic in Early Tests | Newswise

Newswise -- Researchers at McMaster University have developed a new generative artificial intelligence (AI) model capable of drastically speeding up drug discovery -- and, in early tests, it has already designed a brand-new antibiotic. The discovery is a demonstration of how AI could dramatically improve the slow and costly search for new antimicrobial medicines, as bacteria and other microbes continue to evolve resistance to our current suite of drugs. The new model, called SyntheMol-RL, is trained to explore a vast chemical space of up to 46 billion possible compounds -- far beyond what could realistically be tested in the lab, where even large-scale screens top out at around a million molecules. Drawing on roughly 150,000 molecular 'building blocks' and a set of 50 chemical synthesis reactions, the AI model is designed to generate structurally novel antibiotic candidates. "In the lab, we can build chemical compounds using a set of smaller chemical fragments, which can be stuck together like molecular Lego blocks," says Assistant Professor Jon Stokes, whose laboratory developed the new model. "SyntheMol-RL configures those fragments in different ways, faster than humans ever could, to create new, larger chemical compounds that should -- based on its knowledge -- be antibacterial." Stokes, a member of the Michael G. DeGroote Institute for Infectious Disease Research, says that while generative AI is becoming increasingly effective at designing novel antibiotic candidates, key properties that determine a potential drug's clinical viability remain difficult to assess without extensive -- and expensive -- laboratory testing. "It doesn't matter if you find a new chemical that's antibacterial in the lab if it can't dissolve inside the body, if it's toxic to human cells, or if it can't be metabolized and expelled after it has done its job," he explains. "Bleach is antibacterial -- so is fire. But they obviously don't tick those other boxes. Good drug candidates must meet several different criteria, otherwise they'll never become actual medicine." Past iterations of SyntheMol exclusively designed molecules with antibacterial activity, without consideration for these other critical properties. But, over the past two years, Stokes' team -- with collaborators at Stanford University -- has refined the model so it only generates antibacterial compounds that are easy to develop in the lab and likely to be soluble in the body. "There is a lot of conflict between compounds that are antibacterial and compounds that are water soluble," says Gary Liu, a graduate student in Stokes' lab and lead developer of the new model. "In previous studies, filtering for compounds that were both antibacterial and soluble after our prompt often left us with significantly fewer viable drug candidates, so we built solubility right into the generation process and now the model can efficiently design antibiotic candidates with greater clinical promise." In a new study, published April 23 and selected for the cover of the June issue of Molecular Systems Biology, Stokes' team put their enhanced model to the test. They tasked it with generating water-soluble antibiotics that could treat infections caused by Staphylococcus aureus -- colloquially known as "staph infections" -- and quickly got several hits. From a batch of 79 model-proposed antibacterials, Stokes' group homed in on one particularly interesting compound -- a novel, water-soluble compound that seemed likely to have antibiotic activity against S. aureus. The new computer-designed drug candidate, which they called synthecin, was then formulated as a topical cream in the lab and tested on an otherwise drug-resistant wound infection in mouse models. "Synthecin was highly effective at controlling the infection," says Denise Catacutan, a grad student in Stokes' lab who led the wet lab portions of the study. "It worked extremely well as a topical drug, and also shows early promise as something that could be applied or optimized for systemic use in the future." While the new study highlights synthecin's promise, the team has yet to uncover how the drug inhibits bacteria -- a key step, Stokes says, in determining its safety profile and therefore its likelihood of someday landing in clinics. His group is now actively engaged in these critical "mechanism-of-action" studies. Regardless of how those studies play out, though, the group sees the discovery of synthecin as validation that their AI model can rapidly generate high-potential drug candidates, shifting the burden of drug discovery from finding viable compounds to designing and optimizing them. That shift, Stokes says, is significant not only for antibiotic discovery, but also for all areas of biochemistry. "We used our model to design new antibiotics, but it's capable of so much more," says Stokes, a faculty member at the Marnix E. Heersink School of Biomedical Innovation and Entrepreneurship and an executive member of NexusHealth. "We built it to be disease agnostic, meaning it could just as easily generate novel drug candidates for diabetes or cancer or other indications." Stokes' lab continues to enhance SyntheMol and anticipates that an even more robust version will be available later this year. Available for interviews: For any other assistance, contact Adam Ward, media relations officer with McMaster's Faculty of Health Sciences, at [email protected].