AI-redesigned starting points and outcomes enhance protein evolution.

AI-designed enzyme variants improve stability and catalytic efficiency, enabling faster evolution and higher specificity, exemplified by a 79-fold increase for ataxin-2 cleavage.

• Why it matters: Enhancing enzyme stability and activity is crucial for therapeutic and industrial applications, but traditional methods often face limitations in evolving highly functional proteins. Addressing this gap can accelerate enzyme development and expand functional capabilities.
• What they did: Using the ProteinMPNN AI model, three botulinum neurotoxin proteases were redesigned, and their robustness was tested through side-by-side phage-assisted evolution campaigns with wild-type enzymes, totaling four campaigns.
• The result: Redesigned enzymes consistently outperformed wild-type starting points, achieving higher activity, stability, and specificity, notably a 79-fold improvement in ataxin-2 cleavage, demonstrating AI-driven design as a practical tool for enzyme optimization.