The abstract is read at the publisher; the summary is JClub's.
CausalMut improves prediction of enzyme mutational effects on catalytic kinetics across three datasets, reducing root mean square error by 16.24% and increasing R² by 22.19% versus five baselines.
- Why it matters: Existing deep learning models are limited by biases from how enzyme mutation experiments are generated, creating spurious correlations that can hurt generalization under new experimental conditions.
- What they did: CausalMut formalizes the data generation process with a structural causal graph to treat the experimental environment as a confounder, then trains MIENet plus Joint Focal-IPS to correct distributional imbalance.
- The result: In experiments on kcat, Km, and kcat/Km, it shows consistent gains under three distribution-shift settings and case studies that identify high-activity variants within a fixed screening budget.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.