Benchmark Study for Calculations of pKa Values of Metal Ligands in Proteins
Haji Dehabadi, Maryam ; Irani, Mehdi ; Jafari, Sonia and Ryde, Ulf LU
(2026)
In Journal of Chemical Theory and Computation
22(14).
p.7420-7435
- Abstract
We have compared the performance of 64 different computational methods, based on combined quantum mechanical (QM) and molecular mechanical (QM/MM) or QM-cluster calculations in a continuum solvent, to estimate the acid constant (pKa) of metal-bound ligands in proteins. As a calibration set, we use 12 experimental pKa values from six different proteins that involve Zn2+, Fe3+, or Fe4+. We employ two different density functional theory (DFT) methods (TPSS and B3LYP), two basis sets (def2-SV(P) and def2-TZVPD), QM regions of three different sizes (∼40, ∼100, and ∼350 atoms), relaxed or fixed surroundings, and three different values of the dielectric constant of the continuum-solvation model (ε = 4, 20, or 80). The results clearly show that... (More)
We have compared the performance of 64 different computational methods, based on combined quantum mechanical (QM) and molecular mechanical (QM/MM) or QM-cluster calculations in a continuum solvent, to estimate the acid constant (pKa) of metal-bound ligands in proteins. As a calibration set, we use 12 experimental pKa values from six different proteins that involve Zn2+, Fe3+, or Fe4+. We employ two different density functional theory (DFT) methods (TPSS and B3LYP), two basis sets (def2-SV(P) and def2-TZVPD), QM regions of three different sizes (∼40, ∼100, and ∼350 atoms), relaxed or fixed surroundings, and three different values of the dielectric constant of the continuum-solvation model (ε = 4, 20, or 80). The results clearly show that QM-cluster+continuum-solvation is much better than QM/MM. In general, the most accurate results are obtained with ε = 80 and the minimal QM region. The two DFT methods, the two basis sets, and relaxing or fixing the surroundings give similar results. The best-performing method is TPSS with the minimal QM region, def2-TZVPD, relaxed surroundings, and ε = 80, yielding a mean absolute deviation (after removal of a systematic error of 11.6 pKa units, pu) of 2.0 pu and a maximum deviation of 5.0 pu. The coefficient of determination (R2) and Kendall's τ with respect to the experimental pKa values are both 0.64, while Spearman's rank correlation coefficient is 0.78. This level of accuracy should be sufficient to reliably determine the protonation states of metal-bound ligands in QM-based studies of enzymatic reaction mechanisms.
(Less)
- author
- Haji Dehabadi, Maryam
; Irani, Mehdi
; Jafari, Sonia
and Ryde, Ulf
LU
- organization
- publishing date
- 2026-07-28
- type
- Contribution to journal
- publication status
- published
- subject
- in
- Journal of Chemical Theory and Computation
- volume
- 22
- issue
- 14
- pages
- 16 pages
- publisher
- The American Chemical Society (ACS)
- external identifiers
-
- scopus:105046319860
- pmid:42409596
- ISSN
- 1549-9618
- DOI
- 10.1021/acs.jctc.6c00884
- language
- English
- LU publication?
- yes
- id
- 18b6826f-4a6b-42ff-ae51-a8e8bbf57206
- date added to LUP
- 2026-09-28 15:29:13
- date last changed
- 2026-09-29 03:00:02
@article{18b6826f-4a6b-42ff-ae51-a8e8bbf57206,
abstract = {{<p>We have compared the performance of 64 different computational methods, based on combined quantum mechanical (QM) and molecular mechanical (QM/MM) or QM-cluster calculations in a continuum solvent, to estimate the acid constant (pKa) of metal-bound ligands in proteins. As a calibration set, we use 12 experimental pKa values from six different proteins that involve Zn2+, Fe3+, or Fe4+. We employ two different density functional theory (DFT) methods (TPSS and B3LYP), two basis sets (def2-SV(P) and def2-TZVPD), QM regions of three different sizes (∼40, ∼100, and ∼350 atoms), relaxed or fixed surroundings, and three different values of the dielectric constant of the continuum-solvation model (ε = 4, 20, or 80). The results clearly show that QM-cluster+continuum-solvation is much better than QM/MM. In general, the most accurate results are obtained with ε = 80 and the minimal QM region. The two DFT methods, the two basis sets, and relaxing or fixing the surroundings give similar results. The best-performing method is TPSS with the minimal QM region, def2-TZVPD, relaxed surroundings, and ε = 80, yielding a mean absolute deviation (after removal of a systematic error of 11.6 pKa units, pu) of 2.0 pu and a maximum deviation of 5.0 pu. The coefficient of determination (R2) and Kendall's τ with respect to the experimental pKa values are both 0.64, while Spearman's rank correlation coefficient is 0.78. This level of accuracy should be sufficient to reliably determine the protonation states of metal-bound ligands in QM-based studies of enzymatic reaction mechanisms.</p>}},
author = {{Haji Dehabadi, Maryam and Irani, Mehdi and Jafari, Sonia and Ryde, Ulf}},
issn = {{1549-9618}},
language = {{eng}},
month = {{07}},
number = {{14}},
pages = {{7420--7435}},
publisher = {{The American Chemical Society (ACS)}},
series = {{Journal of Chemical Theory and Computation}},
title = {{Benchmark Study for Calculations of pKa Values of Metal Ligands in Proteins}},
url = {{http://dx.doi.org/10.1021/acs.jctc.6c00884}},
doi = {{10.1021/acs.jctc.6c00884}},
volume = {{22}},
year = {{2026}},
}