Regina Barzilay and James Collins have received the 2026 Pritzker Prize for AI in Science Research Excellence for their work on AI-enabled antibiotic discovery. MIT's announcement connects the award to a research program combining deep learning, bioengineering and experimental validation.
What the award recognizes
The prize, established by the University of Chicago with Caltech, recognizes contributions that advance both AI and the natural sciences or engineering. MIT highlights the collaborators' work identifying antibacterial candidates against drug-resistant pathogens.
Their 2020 paper with first author Jonathan Stokes described halicin, a compound previously studied as a possible diabetes drug. MIT reports that a deep-learning model trained on 2,500 molecules identified it and that subsequent experiments tested its antibacterial activity.
Discovery still requires experiments
The announcement also describes the team's 2023 work on abaucin, a narrow-spectrum candidate targeting Acinetobacter baumannii. The contrast between broad and narrow activity illustrates why finding a promising molecule is only one part of a drug-discovery workflow: researchers must investigate what it affects and how it behaves.
The award announcement is not a new treatment approval or evidence that a candidate is ready for routine patient use. It recognizes a body of research. Readers should keep the original studies, their experimental settings and the later development process distinct from the prize itself.
The limits the researchers emphasize
Collins describes a shift toward using generative AI to design molecules as well as identify them. Barzilay cautions that drug discovery remains at an early stage and that AI currently automates parts of the pipeline.
For teams evaluating AI in scientific work, the practical lesson is to inspect the link between a model's proposal and the experiment that tests it. A computational ranking becomes useful evidence when the subsequent measurements support the intended biological claim. The announcement highlights that connection rather than eliminating the need for it.