PAMP: Continuous-to-Discrete Mutation Projection for Enzyme Engineering

PAMP maps a differentiable turnover objective to single-amino-acid substitutions for enzyme engineering, achieving a mean first-round predicted gain of 0.1795 log2 units across 2,754 substrate pairs.

• Why it matters: Enzyme engineering needs a way to convert quantitative turnover forecasts into concrete mutations, reducing the experimental search space while keeping candidate lists small.
• What they did: Using a substrate-conditioned Residual Predictor trained on CataPro-derived data, PAMP applies multi-point gradient averaging and distance-penalized projection in protein embedding space to propose substitutions.
• The result: Positive predicted changes occurred in 84.64% of cases, top-five lists had 99.49% records with at least one positive candidate, and five rounds reached 0.5631 log2 units with consistent method ordering after independent re-scoring.

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