High pKa variability of cysteine residues in structural databases and the effect of H-bond contributions

Abstract

Objective: Our first objective was to extensively compare two most common empirical protein pKa predictors, propka1.0 (ppka1) and propka3.0 (ppka3); we have specifically compared them as tools for high-throughput analyses of structural datasets with a particular focus on the amino acid Cysteine (Cys); afterwards, our goal was to assess their performances with known instances of reactive Cys residues. Methods: Structural datasets were downloaded from the PDB repository and pipelined to different pKa prediction software; results were parsed with in-house scripts, to extract relevant information, and then subjected to further analysis, including detailed output comparisons for different programs. Results: With ppka1, H-bond contributions dictated the prediction of Cys pKa, particularly for exposed residues; this was not the case for the most recent version, ppka3. This feature of ppka1 fits with recent, independent studies reporting the critical role of H-bond network in the activation of reactive Cys residues; indeed, when tested in a benchmark for its ability to describe reactive Cys residues, ppka1 provided the best results, favorably comparing to other methods tested. Conclusion: ppka1 can be an effective aid in redox bioinformatics as a tool for high-throughput Cys pKa predictions: it is extremely fast, yet capable of competitive performances, particularly apt to predict very reactive (e.g. nucleophilic, exposed) functional Cys residues. This work provides new insights on propka (in its different versions) predictions as well as substantial support to the critical role of H-bond and exposure in Cys activation.