Unraveling the energetic significance of chemical events in enzyme catalysis via machine-learning based regression approach

Bacterial β-lactamases, enzymes contributing to antibiotic resistance, are studied using hybrid quantum mechanical molecular mechanical (QM/MM) chain-of-states pathway optimizations. Minimum energy pathways are sampled via re-optimization under various protein environments from constrained molecular dynamics. Machine learning regression creates predictive potential energy surface models. Two model-independent criteria quantify energetic contributions and correlations, identifying rate-limiting steps. The workflow's consistency is tested across various quantum chemistry theories and machine-learning models, showing qualitative agreement with experimental mutagenesis studies. The TEM-1/benzylpenicillin acylation reaction serves as a model system.

Zilin Song, Hongyu Zhou, Hao Tian, Xinlei Wang, Peng Tao