Scientists at the University of Washington, led by David Baker, used AI‑based design tools to build new proteases that cut chosen protein sequences. They started from a zinc protease and a cysteine protease, redesigning the rest of each enzyme with RFdiffusion2 and ProteinMPNN. After testing dozens of variants, some designs cut target proteins in cells, including proteins tied to Alzheimer’s and frontotemporal dementia. The best versions were up to 10‑fold more active than the original designs.
Why it matters
If refined, such AI‑designed enzymes could become tools for treating protein‑misfolding diseases.
