AI tool shrinks proteins by 10–15% while preserving function

Raygun represents a distinct strategy in protein engineering: rather than designing proteins entirely from scratch, it works within the boundaries of known protein sequences, using AI to identify which amino acids are truly essential and which can be removed or substituted. This approach, led by computational biologist Rohit Singh at Duke University, harnesses evolutionary patterns embedded in protein sequences to guide the search for critical components—a method that allows for more extensive modifications than traditional point-by-point mutation testing would suggest.

The practical implications align with real bottlenecks in biotechnology. Gene therapy faces a fundamental constraint: viral delivery vehicles have limited carrying capacity, and any reduction in the size of a therapeutic protein could expand the range of candidates. Similarly, fluorescent protein tags, while invaluable for imaging biological processes, can sometimes alter the very systems they are meant to observe; miniaturizing them could mitigate this interference. Singh's team views Raygun not as a replacement for de novo design but as a complementary tool that can operate on top of newly engineered proteins, effectively allowing researchers to refine and optimize novel designs further.