EZSolver: Template-Free Prediction of Polar Enzymatic Mechanisms via Bidirectional Flow Matching and Search

Abstract Predicting enzymatic reaction mechanisms is critical for understanding enzyme function and for the design and discovery of new enzymes. Current computational predictors rely on deterministic, rule-based dictionaries, which perform well on in-distribution tasks but fail to generalize to out-of-distribution (OOD) chemistry. To address this limitation, we present EZSolver for the polar enzymatic mechanism prediction. Powered by a flow-matching predictor (EZFlow), a template-free, generative framework, and navigated by an evaluator-guided bidirectional beam search, EZSolver learns the chemistry of electron redistribution instead of memorizing rigid templates. Evaluated across diverse enzyme classes, EZSolver achieves a 60.0% accuracy and an 84.6% chemical plausibility rate for the full-mechanism prediction of unseen polar enzymatic reactions. While rule-based models collapse without predefined templates, EZSolver successfully extrapolates chemical knowledge to infer uncatalogued pathways, as demonstrated during rigorous OOD benchmarking. By illuminating enzymatic chemical mechanisms, EZSolver helps pave the way for the automated prediction of enzyme function and the discovery and design of novel biocatalysts for sustainable chemistry.