Electrostatic pKa computations in proteins: role of internal cavities

The solvent accessible surface area (SASA) algorithm is conventionally used to characterize protein surfaces in electrostatic energy computations of proteins. Unfortunately, it often fails to find narrow cavities inside a protein. As a consequence pK(a) computations based on this algorithm perform badly. In this study a new cavity-algorithm is introduced, which solves this problem and provides improved pK(a) values. The procedure is applied to 20 pK(a) values of titratable groups introduced as point mutations in SNase variants, where crystal structures are available. The computations of these pK(a)s are particular challenging, since they are placed in a rather hydrophobic environment. For nine mutants, where the titratable residue is in contact with a large cavity, the RMSD(pKa) between computed and measured pK(a) values is 2.04, which is a considerable improvement as compared to the original results obtained with Karlsberg(+) (http://agknapp.chemie.fu-berlin.de/karlsberg/) that yielded an RMSD(pKa) of 8.8. However, for 11 titratable residues the agreement with experiments remains poor (RMSD(pKa) = 6.01). Considering 15 pK(a)s of SNase, which are in a more conventional less hydrophobic protein environment, the RMSD(pKa) is 2.1 using the SASA-algorithm and 1.7 using the new cavity-algorithm. The agreement is reasonable but less good than what one would expect from the general performance of Karlsberg(+) indicating that SNase belongs to the more difficult proteins with respect to pK(a) computations. We discuss the possible reasons for the remaining discrepancies between computed and measured pK(a)s.