Relative Enzymatic Activity Levels From In Silico Mutagenesis.

We propose a new computational approach for predicting the impact of point mutations on residual enzymatic activities. We build on the general linear trends existing between free energy and enthalpy of transfer of substrates from cytosol to enzyme active sites (protein-ligand binding), therefore linking the docking energies to the binding free energies. In this very first step, we rationalize these trends in terms of a compensation effect decomposed into explicit thermodynamics contributions. In a second step, we combine the latter with the assumption that free energies of transfer, estimated from docking, and free energies of activation are linearly related through a Bronsted-Evans-Polanyi (BEP) relationship, allowing us in fine to predict enzyme activity. As a result, we propose generic Langmuir-Hinshelwood kinetic equations "trained" on the wild type, which provide excellent predictions of rates of catalytic transformations for mutated enzymes from the combination of in silico docking energies to a set of system-specific experimental data. This generalized approach is validated against clinical data on the particular case of human fumarase with major implications for the understanding of hereditary fumarase deficiency.