EnzOracle: Mechanism-aware prediction of enzyme environmental adaptation via a classification-guided mixture-of-experts framework

Industrial biocatalysis increasingly requires enzymes capable of operating under extreme physicochemical conditions, yet most natural sequence data reflect adaptation to mild environments, leading conventional predictive models to suffer from regression-to-the-mean effects in extremophilic regimes. Here we present EnzOracle, a classification-guided mixture-of-experts framework that enables distribution-aware prediction of enzyme melting temperature ( T _m ), optimal catalytic temperature ( T _opt ), and optimal pH ( pH _opt ) directly from sequence. EnzOracle demonstrated robust predictive accuracy across diverse benchmarks, achieving RMSE of 5.245°C for T _m , 11.458°C for T _opt , and 0.781 for pH _opt . Beyond predictive accuracy, we introduce a trait-resolved molecular simulation strategy to evaluate whether EnzOracle-derived attribution patterns correspond to independent physical mechanisms. Across representative systems, attention hotspots mapped onto rigidity-conferring interaction networks for T _m , dynamically preorganized active-site ensembles for T _opt , and pH-dependent electrostatic and hydration networks for pH _opt . These orthogonal validations indicate that EnzOracle captures transferable biophysical principles of enzyme environmental adaptation rather than merely exploiting dataset-specific correlations, positioning sequence-based learning as a mechanism-aware framework for discovering stability and activity determinants across diverse catalytic landscapes.