BU researchers showed that an AI model built around the biology of antibodies can outperform much larger general models at predicting antibody designs. The approach could shorten the early stages of drug discovery.