AI-selected mutations sharply increase bacterial virus infection in lab tests

University of Wisconsin-Madison researchers used an AI model to identify mutations that made engineered viruses infect bacterial hosts up to six orders of magnitude more effectively than their naturally occurring counterparts in laboratory tests. The viruses, called phages, infect bacteria and are being studied as possible alternatives to antibiotics. The team trained its model on experimental data showing how changes to phage proteins affect infection. It also used the model to identify mutations aimed at particular bacteria while sparing others. The result demonstrates a way to search for more effective phages, but the reported gains come from laboratory evaluations; the source does not report a treatment tested in patients.