pKa prediction in non-aqueous solvents

Abstract

Acid dissociation constants (pKa ) are widely measured and studied, most typically inwater. Comparatively few datasets and models for non-aqueous pKa values exist. Inthis work, we demonstrate how the pKa in one solvent can be accurately determinedusing reference data in another solvent, corrected by solvation energy calculationsfrom the COSMO-RS method. We benchmark this approach in 10 different solvents,and find that pKa values calculated in six solvents deviate from experimental data onaverage by less than 1 pKa unit. We observe comparable performance on a morediverse test set including amino acids and drug molecules, with higher error for largemolecules. The model performance in four other solvents is worse, with one MAEexceeding 3 pKa units; we discuss how such errors arise due to both model error andinconsistency in obtaining experimental data. Finally, we demonstrate how this tech-nique can be used to estimate the proton transfer energy between different solvents,and use this to report a value of the proton's solvation energy in formamide, a quan-tity that does not have a consensus value in literature.