A research team has turned to Google's AlphaFold, the AI system built to predict protein structures, to redesign gene-editing proteins and make them safer, according to Ars Technica.
Gene-editing tools work by using proteins to target and alter specific stretches of DNA. But those proteins don't always land precisely where intended. Off-target activity — edits made in the wrong place — has long been one of the central safety concerns for the field, since unintended changes to a genome can carry unpredictable consequences.
According to Ars Technica, the team used AlphaFold to help identify which parts of a gene-editing protein are responsible for those mistakes. Pinpointing the structural regions tied to errant behavior gives researchers a target: with that map in hand, they can redesign the protein to cut down on the errors while preserving its intended function.
The approach reflects a broader shift in biology, where AI models that predict how proteins fold and behave are increasingly used not just to understand existing proteins but to engineer better ones. Ars Technica frames the work as using AlphaFold to ID the problem regions and then redesign around them.
Why it matters: safer, more precise gene-editing proteins could reduce the risk of unintended DNA changes, a key hurdle standing between experimental editing techniques and their wider use in medicine and biotechnology.