Accurate Prediction of Noncovalent Interaction Energies with the Effective Fragment Potential Method: Comparison of Energy Components to Symmetry-Adapted Perturbation Theory for the S22 Test Set.

Noncovalent interactions play an important role in the stabilization of biological molecules. The effective fragment potential (EFP) is a computationally inexpensive ab initio-based method for modeling intermolecular interactions in noncovalently bound systems. The accuracy of EFP is benchmarked against the S22 and S66 data sets for noncovalent interactions [Jurečka, P.; Šponer, J.; Černý, J.; Hobza, P. Phys. Chem. Chem. Phys.2006, 8, 1985; Řezáč, J.; Riley, K. E.; Hobza, P. J. Chem. Theory Comput.2011, 7, 2427]. The mean unsigned error (MUE) of EFP interaction energies with respect to coupled-cluster singles, doubles, and perturbative triples in the complete basis set limit [CCSD(T)/CBS] is 0.9 and 0.6 kcal/mol for S22 and S66, respectively, which is similar to the MUE of MP2 and SCS-MP2 for the same data sets, but with a greatly reduced computational expense. Moreover, EFP outperforms classical force fields and popular DFT functionals such as B3LYP and PBE, while newer dispersion-corrected functionals provide a more accurate description of noncovalent interactions. Comparison of EFP energy components with the symmetry-adapted perturbation theory (SAPT) energies for the S22 data set shows that the main source of errors in EFP comes from Coulomb and polarization terms and provides a valuable benchmark for further improvements in the accuracy of EFP and force fields in general.