AI-designed phage mutations boost bacterial infection up to a millionfold

Researchers at the University of Wisconsin–Madison have used AI to identify phage mutations that enable viruses to infect bacteria up to one million times more effectively, a finding published in Cell Systems that could accelerate alternatives to traditional antibiotics. The AI model, developed by biochemistry professor Vatsan Raman's team, was trained on tens of thousands of lab-generated mutations to learn rules for phage success. Tests showed engineered phages with AI-suggested mutations outperformed natural counterparts by up to six orders of magnitude, while also enabling targeting of specific bacteria. This work builds on prior research into phage proteins that act as molecular drills, and the team is now applying the approach to pathogens causing urinary tract infections, microbiome-linked diseases, and Salmonella in poultry, aiming to design therapeutic phages for diverse bacterial threats.