Assessing Alkene Reactivity toward Cytochrome P450-Mediated Epoxidation through Localized Descriptors and Regression Modeling

The prediction of sites of epoxidation by cytochrome P450s during metabolism is particularly important in drug design, as epoxides are capable of alkylating biological macromolecules. Reliable methods are needed to quantitatively predict P450-mediated epoxidation barriers for inclusion in high-throughput screening campaigns alongside protein-ligand docking. Utilizing the fractional occupation number weighted density (FOD) and orbital-weighted Fukui index (f_w⁺) as descriptors of local reactivity and a data set of 36 alkene epoxidation barriers computed with density functional theory (DFT), we developed and validated a multiple linear regression model for the reliable estimation of epoxidation barriers using only substrate structures as input. Using our recommended level of theory (GFN2-xTB//GFN-FF), mean absolute errors in the training and test sets were found to be 0.66 and 0.70 kcal/mol, respectively, with coefficients of determination of ca. 0.80. We demonstrate the utility of this approach on three known substrates of CYP101A1 and further show that this approach is inappropriate for particularly electron-rich alkenes. By employing a modern semiempirical method on force-field-generated geometries, the required descriptors can be calculated on the millisecond timescale per structure, making the approach well suited for incorporation into high-throughput methodologies alongside docking.