(652c) Quick and Accurate Estimates of Mutation Effects on Relative Activity of Enzymes from Molecular Simulations with Restrained Transition States

Many of the most promising approaches in enzyme engineering from the past several years rely on the collection and analysis of large libraries of enzyme variants, either to directly search for improved variants as in directed evolution, or more recently to train machine learning models to infer relationships between sequence and structure or function. However, despite ever-increasing computational power, it usually remains infeasible to run a direct atomistic simulation for each of a large body of variants. This problem is further exacerbated for studies of rare events like chemical reactions, which necessitate the use of expensive enhanced sampling methods, and for especially large proteins or protein complexes. These factors have heretofore limited the applicability of atomistic simulations in data science approaches for enzyme engineering. Here we present a novel but highly simple approach to comparing enzyme variants with fully atomistic classical molecular dynamics simulations on a tractable timescale. Our method greatly simplifies the problem by restricting sampling only to the reaction transition state, and we show that it can be used to measure the relative effects of mutations with very high accuracy across two distinct enzymes, even where free energy methods would be unable to do so. This method will enable atomistic simulations to achieve sampling coverage for enzyme variant screening on a scale that was previously only possible for experimental methods or highly simplified models.