toward Chemical-Group-Independent p K a Estimation for Pharmaceutical Ingredients and Biomolecules
We investigated whether the systematic errors of conventional Gibbs energy difference (ΔG)-based aqueous pKa estimation can be interpreted and reduced using multivariate analysis of quantum-chemical descriptors. Ten descriptors obtained from quantum chemical calculations, including the deprotonation Gibbs energy difference, local solvation terms, proton charge, and donor–proton bond length, were analyzed for 427 compounds. A single model covering all donor elements was not satisfactory; however, donor specific models substantially reduced the systematic residuals of a ΔG-based linear calibration. The fitted models gave coefficients of determination as high as R2 = 0.97 and mean absolute errors of approximately 0.5–0.6 pKa units for the O–H and N–H data sets. Applications to representative pharmaceutical ingredients reproduced the experimentally assigned order of deprotonation in several polyprotic molecules and generally gave deviations within approximately 1 pKa unit. Applications to nucleobases, metal-coordinated nucleobases, persulfides, cysteine, and a green fluorescent protein chromophore model further showed that the framework can support qualitative and semiquantitative comparisons across chemically diverse systems. This model on S–H was less robust because of the smaller data set and evidence of overfitting; for these compounds, direct ΔG-based estimations were often similarly reliable. The results show that ΔG remains the dominant determinant of pKa, whereas local structural descriptors just account for systematic deviations that cannot be represented by a single ΔG-based calibration line.
Authors
- Toru Matsui (ORCID: https://orcid.org/0000-0001-9847-5925)
Institutions
- University of Tsukuba (JP)
Publication Details
- Journal
- The Journal of Physical Chemistry A
- Published
- 2026-11-19
- DOI
- https://doi.org/10.1021/acs.jpca.6c05243
- Primary Topic
- Advanced Chemical Physics Studies
- Type
- article
- Field-Weighted Citation Impact
- 0.00