AI-designed proteases target neurological proteins for chronic disease treatment

University of Washington researchers created AI-guided proteases that cleave specific disease proteins, potentially lowering costs for neurological conditions

University of Washington researchers used AI tools—RFdiffusion2 and ProteinMPNN—to design proteases targeting neurological disease proteins. The approach generated 12 cysteine protease designs with 69 amino acid configurations, achieving 19% functional cleavage in initial lab tests using fluorescent protein markers. A separate set of 135 zinc protease designs showed no structural matches in existing databases (AlphaFold3), though these were not validated for in vivo use. The work builds on zinc proteases from crayfish digestive systems and cysteine proteases from yeast, aiming to disrupt proteins previously considered unbreakable by conventional methods. This could reduce costs for chronic neurological conditions where current treatments are expensive and limited. What matters next: independent validation of in vivo efficacy for zinc proteases and scaling the 19% success rate to real-world biological systems. The source does not provide details on clinical applications or cost reductions. The 19% success rate refers to functional cleavage in initial test conditions with fluorescent protein markers, and zinc protease designs were not fully validated for in vivo efficacy.

Source: Ars Technica