Researchers use AI to design proteins that cut up other proteins

Key Points:

• Researchers at the University of Washington used AI-based protein design tools to engineer novel proteases by retaining catalytic cores from existing enzymes and redesigning the rest of the protein, enabling precise cleavage of targeted proteins.
• The team employed RFdiffusion2 to generate enzyme backbones that position key catalytic amino acids and ProteinMPNN to assign compatible amino acids, producing unique protease designs not found in natural protein databases.
• Experimental testing showed a subset of these AI-designed proteases successfully cleaved target proteins, with activity improved significantly through targeted mutations near the active site, achieving up to 10^9-fold acceleration in peptide bond cleavage.
• The researchers demonstrated therapeutic potential by designing zinc proteases capable of cutting proteins implicated in neurological diseases like Alzheimer's and frontotemporal dementia within cells.
• Despite promising results, challenges remain for clinical application, including delivery to target sites, immune responses, and the current AI tools' limited ability to optimize catalytic activity beyond structural constraints.