The prediction of biochemical acid dissociation constants using first principles quantum chemical simulations

Proton transfer is a vital part of many chemical processes and is determine d by the acid dissociation constants (pK(a)) of the chemicals involved. The goal of this study is to evaluate existing quantum chemical methods to acc urately predict pK(a) and to determine the trade-off between accuracy and c omputational cost. We used density functional theory (DFT) with the B3LYP f unctional and two basis sets, 6-31G** and 6-31++G(3df,3pd). To include the effects of aqueous solvation, we used DFT combined with a polarizable conti nuum solvation model as well as the Langevin dipole model. Using largest ba sis set and aqueous-phase optimized structures, we find a strong linear rel ationship between the predicted and calculated pK(a)'s R-2 = 0.94 and even stronger fits within the individual classes of compounds. Despite these str ong linear correlations, we find a systematic error leading to a much large r range in the predicted pK(a) values. We studied the effect of excluding t he enthalpy and entropy terms, using the gas-phase optimized structures, an d using a smaller basis set, and found in all cases no significant decrease in the accuracy of the predicted pK(a) values. We also calculated the pK(a ) values from the gas-phase energies, excluding any solvation effects, and also found a strong linear relation with the experimental pK(a) values, alt hough with much larger systematic errors in the range of predicted pK(a) va lues. Finally, we found that the Langevin dipole method yielded pK(a) value s with smaller absolute errors than the PCM methods, bur yielded poorer lin ear fits to the experimental values. (C) 2000 Elsevier Science B.V. All rig hts reserved.