Assessment of Methodology and Chemical Group Dependences in the Calculation of the pKa for Several Chemical Groups.

We have investigated the dependencies of various computational methods in the calculation of acid dissociation constants (pKa values) of certain chemical groups found in protonatable amino acids based on our previous scheme [ Matsui ; Phys. Chem. Chem. Phys. 2012 , 14 , 4181 - 4187 ]. By changing the quantum chemical (QC) method (Hartree-Fock (HF) and perturbation theory, and composite methods, or exchange-correlation functionals in density functional theory (DFT)), basis sets, solvation models, and the cavities used in the solvent models, we have exhaustively tested about 2,200 combinations to find the best combination for pKa estimation among them. Of the tested parameters, the choice of the basis set and cavity is the most crucial to reproduce experimental values compared to other factors. Concerning the basis set, the inclusion of diffuse functions is quite important for carboxyl, thiol, and phenol groups judging from the mean absolute errors (MAEs) measured from the experimental values. Of the cavity models, between the Pauling, Klamt, and the universal force field (UFF) definitions, the UFF defined cavity is the best choice, resulting in the smallest MAEs. Concerning the QC methods, hybrid DFTs and range-separated DFTs always provide better results than pure DFTs and HF. As a result, we found that LC-ϖPBE/6-31+G(d) with PCM-SMD/UFF provides the best pKa estimation with a MAE within 0.15 pKa units.