Identifying beneficial enzyme mutations for non-native chemical transformations is key to biocatalyst optimization. However, the unknown relationships between enzyme structure and function hinder the efficient discovery of beneficial enzyme mutants. Molecular simulations, including quantum mechanics and molecular mechanics, have been extensively applied to evaluate the impact of the mutation on free energy barriers. But these simulations are typically resource and time-consuming and are hard to incorporate in high-throughput virtual protocols. To address this challenge, we have been developing simulation-derived, computationally-inexpensive physical descriptors that can qualitatively reflect the impact of mutations on enzyme reactivity and selectivity. In this work, we will discuss the search for the predictive physical descriptors using a lactonase <i>Sso</i>Pox as a model system. By combining molecular dynamics simulations and machine learning methods, we have identified a descriptor for enzyme/substrate fitness: solvent accessible surface area ratio between substrate and enzyme (SASA ratio). SASA ratio shows a volcano-type correlation with the activation free energy of enzyme mutants. It can also explain the alteration of substrate specificity upon mutations.

Identifying descriptors for predicting kinetic mutational effects of lactonase by integrating machine learning and atomistic simulations

ACS Spring 2022

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.