To tackle the challenge of treating drug-resistant bacterial infections, MIT researchers Regina Barzilay and James J. Collins have collaborated on AI-accelerated antibiotic R&D, earning them the 2026 Pritzker Award for Excellence in AI Science Research—the first time this award has gone to a collaborative research team. Their project has completed virtual screening of more than 70 billion molecules and antibacterial testing on over 38,000 molecules, yielding a cumulative total of nine new antibiotic candidates, whether discovered or designed. The work began with using AI to screen existing compound libraries, which successively led to the discovery of completely novel candidate antibiotics: Halicin and Abaucin, the latter targeting Acinetobacter baumannii. Through interpretable deep learning models, the team also identified molecular structural features associated with antibacterial activity. They then moved on to directly constructing entirely new molecular structures using generative AI, producing active candidates such as NG1, effective against drug-resistant Neisseria gonorrhoeae, and DN1, effective against multidrug-resistant Staphylococcus aureus. The team has further built a closed-loop R&D pipeline that links computational prediction, molecular design, and experimental validation. Currently, they are focusing on seven key pathogenic bacteria while advancing candidates toward practical drug development. However, turning active molecules into usable antibiotics still requires clearing stages including synthesis validation and clinical development, as well as resolving challenges related to data quality, model evaluation, and regulation, among other aspects.