Enzyme engineering is an increasingly important field in biotechnology, industry, and medicine. Traditionally, new enzymes are developed through directed evolution, an iterative trial-and-error process that requires substantial time and experimental effort[1]. Computational enzyme design offers an alternative by predicting variants with desired properties. While thermodynamic properties such as stability can often be estimated using classical molecular mechanics (MM) simulations, catalytic activity remains considerably more challenging because it requires accurate evaluation of reaction mechanisms. Consequently, studies of enzyme activity typically rely on quantum mechanical (QM) or QM/MM methods, whose computational cost limits the number of variants that can be screened[2].

Here, we introduce a methodology based on non-equilibrium alchemical transformations to predict mutation-induced changes in activation free energies \(\Delta \Delta G^{\ddagger}\). Because the approach employs a purely classical molecular description, it is at least an order of magnitude faster than conventional QM/MM calculations. To obtain bonded parameters, we developed a new scheme combining the inversed Hessian matrix of the transition state and the modified Seminario method [3]. Non-bonded parameters are derived from substrate electron densities using the atoms-in-molecules method [4]. The proposed approach is tested in two enzymatic systems: Crotonyl-CoA Carboxylase/Reductase and Dihydrofolate Reductase. The results show that the alchemical transformations successfully reproduce both experimental values and trends, with errors close to the chemical precision (1 kcal/mol)[5]. Finally, we applied our workflow to an Old Yellow Enzyme variant that catalyzes C–C bond formation via an aldol reaction. The method identified a novel double mutant (Y187F/T130A) with a twofold increase in activity relative to the native enzyme. Experimental validation showed good agreement with the computational predictions, demonstrating the accuracy of the workflow and its potential for high-throughput enzyme engineering.

References
[1] F. H. Arnold, Angewandte Chemie, 2018, 57, 4143–4148.
[2] V. Vennelakanti, A. Nazemi, R. Mehmood, A. H. Steeves and H. J. Kulik, Current Opinion in Structural Biology, 2022, 72, 9–17.
[3] A. E. A. Allen, M. C. Payne and D. J. Cole, Journal of Chemical Theory and Computation, 2018, 14, 274–281.
[4] F. Heidar-Zadeh, P. W. Ayers, T. Verstraelen, I. Vinogradov, E. Vohringer-Martinez and P. Bultinck, J. Phys. Chem. A, 2018, 122, 4219–4245.
[5] C. Castillo-Orellana and E. Vohringer-Martinez, Chemical Communications, 2025, 61,16078–16081.