Harnessing catalytic knowledge to guide enzyme discovery and evolution with AlphaEnzyme

AlphaEnzyme, a structure-grounded enzyme–reaction foundation model, reaches 30.1% top-1 catalyst retrieval success on a time-separated benchmark, far above a 17.2% baseline.

• Why it matters: Improving enzymes requires linking evolutionary signals to the structural and chemical context that governs catalysis, kinetics, and how mutations respond.
• What they did: AlphaEnzyme jointly reads catalytic compatibility, Michaelis constant, and mutational response from an enzyme paired with its reaction, and is evaluated on retrieval, kinetics rank correlation, and prospective engineering.
• The result: It attains a Michaelis constant rank correlation of 0.64 for enzymes sharing ≤40% identity with the kinetic training set, enriches high-fitness TrpB variants within 16 candidates, and yields a lysate-assay design with a 30.4-fold higher product signal while retaining (R)-selectivity.

AI summary of the abstract · may contain errors · not medical advice · read the paper.