(544eb) Exploring Biocatalyst Design and Process Optimization Using Active Learning and Atomistic Simulations

The industrial application of enzymes, especially in detergent and food industries, is a fast growing sector with a promise of a more sustainable green future. However, studies on the modification of enzymes and their respective catalytic processes for the non-biomedical field is comparatively small with respect to biomedical applications, such as drug development. A powerful tool to contribute to these studies on catalytic activity is the computational modeling of enzyme function, particularly at the atomistic scale. There have been a variety of atomistic modeling tools that have shown promise for the study of enzymatic catalysis, including quantum mechanical (QM) cluster, hybrid quantum mechanical(self-consistent field)/molecular mechanics (QM(SCF)/MM), and hybrid QM(EVB)/MM methodologies. The empirical valence bond (EVB) method is a hybrid QM/MM method that describes reactions by mixing diabatic states that correspond to classical valence bond structures, which represent the reactant, intermediate (or intermediates), and product states¹. Here we will focus upon the QM(EVB)/MM methodology, where key force field parameters are estimated from experimental values or QM calculations. Active learning concepts were used to develop our initial model parameterization and will be discussed.

Our results will focus upon the computational study of serine protease activity. In this study, the Gibbs free energy of activation (Δg^‡_cat), rate coefficient (k), and the Gibbs free energy of reaction (ΔG_rxn) were calculated using QM(EVB)/MM and free-energy perturbation (FEP)/umbrella sampling methods, and these values were then compared with experimental values. With these key thermochemical and kinetic parameters, detailed reaction path analysis was performed on peptide bond cleavage, and Gibbs free energy surfaces of reaction were developed for reaction in water and in the native enzyme. Furthermore, our studies explored key environmental variables affecting catalysis including temperature and pH, which can be used in process optimization for commercialized enzymes or for the scale-up of promising enzyme candidates. Key amino acid residues in the native enzyme active site were mutated to determine overall effects on catalytic activity. These learnings will be discussed in the context of enzyme design.

1. Adamczyk, A. J.; Warshel, A., Converting structural information into an allosteric-energy-based picture for elongation factor Tu activation by the ribosome. Proc. Natl. Acad. Sci. U. S. A. 2011, 108 (24), 9827-9832.