A fast and interpretable deep learning approach for accurate electrostatics-driven pKa predictions in proteins

Abstract Existing computational methods to estimate pKa values in proteins rely on theoretical approximationsand lengthy computations. In this work, we use a data set of 6 million theoretically determined pKashifts to train deep learning models that are shown to rival the physics-based predictors. Theseneural networks managed to assign proper electrostatic charges to chemical groups, and learned theimportance of solvent exposure and close interactions, including hydrogen bonds. Although trained onlyusing theoretical data, our pKAI+ model displays the best accuracy on a test set of ∼750 experimentalvalues. Inference times allow speedups of more than 1000 times faster than physics-based methods.By combining speed, accuracy and a reasonable understanding of the underlying physics, our modelsprovide a game-changing solution for fast estimations of macroscopic pKa from ensembles of microscopicvalues as well as for many downstream applications such as molecular docking and constant-pHmolecular dynamics simulations.