The interaction energies of ubiquitous weakly polar interactions in proteins are comparable with those of hydrogen bonds, consequently, they stabilize local, secondary, and tertiary structures. However, the most widely-used density functionals fail to describe the weakly polar interactions. Thus, it is important to find and test functionals which adequately describe and quantify the energetics of such interactions. For this purpose, interaction energies in the hydrophobic core of rubredoxin (PDB id: 1rb9) and in the S22 subset of the JSCH-2005 benchmark database were computed with the BHandHLYP and PWPW91 functionals and with the pseudospectral implementation of the local MP2 (PSLMP2) method. The cc-pVDZ, cc-pVTZ(-f), cc-pVTZ, cc-pVQZ(-g), aug-cc-pVDZ, aug-cc-VTZ(-f), and aug-cc-pVTZ basis sets were used for the calculations. In the S22 subset the PS-LMP2 results were extrapolated to the complete basis set limit. Furthermore, the a posteriori counterpoise method of Boys and Bernardi was used to correct the basis set superposition errors in the calculation of interaction energies. Calculations using the BHandHLYP functional, both for the various weakly polar interactions in rubredoxin and for the dispersion interactions in the S22 subset, were in good agreement with those using the coupled cluster (CCSD(T)) and the resolution of identity MP2 (RIMP2) methods and clearly outperformed both the PWPW91 functional and the PS-LMP2 method. The results for the S22 hydrogen bonded subset, obtained with PWPW91 calculations, were closest to those of the reference high level calculations. For the "mixed" (hydrogen bonded and dispersive) interactions in the S22 subset, results obtained with the BHandHLYP and PS-LMP2 calculations agreed well with the reference calculations.
All Science Journal Classification (ASJC) codes
- Computational Mathematics