Benchmarking In Silico Tools for Cysteine pK_a Prediction

Abstract

Accurate estimation of the pK a’s of cysteine residues in proteins could inform targeted approaches in hit discovery. The pK a of a targetable cysteine residue in a disease-related protein is an important physiochemical parameter in covalent drug discovery, as it influences the fraction of nucleophilic thiolate amenable to chemical protein modification. Traditional structure-based in silico tools are limited in their predictive accuracy of cysteine pK a’s relative to other titratable residues. Additionally, there are limited comprehensive benchmark assessments for cysteine pK a predictive tools. This raises the need for extensive assessment and evaluation of methods for cysteine pK a prediction. Here, we report the performance of several computational pK a methods, including single-structure and ensemble-based approaches, on a diverse test set of experimental cysteine pK a’s retrieved from the PKAD database. The dataset consisted of 16 wildtype and 10 mutant proteins with experimentally measured cysteine pK a values. Our results highlight that these methods are varied in their overall predictive accuracies. Among the test set of wildtype proteins evaluated, the best method (MOE) yielded a mean absolute error of 2.3 pK units, highlighting the need for improvement of existing pK a methods for accurate cysteine pK a estimation. Given the limited accuracy of these methods, further development is needed before these approaches can be routinely employed to drive design decisions in early drug discovery efforts.

Accurately Predicting Protein pKa Values Using Nonequilibrium Alchemy

“…While our results and the recent work of others , underscore the p K a prediction accuracy attainable by MD-based free energy methods, the gap between prediction and experiment remains larger than the experimental error of 0.1–0.2 p K units . In the current work, we have identified two main sources contributing to the p K a prediction error: residue coupling and force-field parametrization.…”

“…Our finding underscores the importance of accurately parametrizing both the protonated and deprotonated forms of the amino acids and the sensitivity that relative free energy calculations can have to seemingly minor parametrization differences. Suggestive of this phenomenon was the recent demonstration that modification of the Amber14SB cysteine thiolate parametersto agree more closely with ab initio solvation datacould improve the p K a prediction accuracy by 0.5 p K units when combined with an MD-based approach . The use of polarizable force fields might also improve p K a estimates; however, recent work using Monte Carlo simulations with the Drude force field and a Poisson–Boltzmann continuum solvent model did not show a significantly improved prediction accuracy …”

Molecular Insights into the Impact of Mutations on the Binding Affinity of Targeted Covalent Inhibitors of BTK

“…TCIs bind to a nucleophilic amino acid side chain of a target protein that is implicated in a disease condition. Cysteines have typically been the nucleophile of choice due to the nucleophilicity of its thiol side chain (−SH). − However, more recent efforts have focused on targeting nucleophiles beyond cysteines (e.g., lysine) − in an effort to diversify and expand the landscape of druggable residues for covalent modification.…”

Landscape of In Silico Tools for Modeling Covalent Modification of Proteins: A Review on Computational Covalent Drug Discovery

“…A p K a study was done by Awoonor-Williams et al on a diverse test set of experimental cysteine p K a ’s retrieved from the PKAD database to test the performance of several computational p K a methods. This is important, since the fraction of nucleophilic thiolate amenable to chemical modification in a disease-related protein is influenced by the p K a of a targetable cysteine residue while traditional structure-based in silico tools are limited in their predictive accuracy of cysteine p K a ’s relative to other titratable residues.…”