MIT researchers Regina Barzilay and James Collins have won the 2026 Pritzker Prize for AI in Science Research Excellence. The award recognises their joint work combining artificial intelligence with biological research and experimental testing to discover antibiotics against drug-resistant bacteria, which can survive existing treatments.
Their first major paper together, published with Jonathan Stokes in 2020, described an antibiotic called halicin. A deep learning system, trained on 2,500 molecules, identified the substance, which had previously been studied as a possible diabetes drug. Tests showed that halicin could kill several kinds of antibiotic-resistant bacteria. The discovery took weeks rather than years.
In 2023, the researchers worked with Stokes again to identify abaucin. Unlike halicin, which acts against a broad range of bacteria, abaucin targets a single species. This focused action reduces the likelihood of bacteria developing resistance and limits the risk of harming useful bacteria in the gut.
Collins says traditional antibiotic discovery has involved slow, expensive and unpredictable trial and error. He sees generative AI changing its role from finding promising substances to designing them. However, Barzilay stresses that AI currently automates only parts of drug discovery. She says more realistic models of human biology are needed to transform the whole process.
