A systematic assessment of density functionals and ONIOM schemes for the study of hydrogen bonding between water and the side chains of serine, threonine, asparagine, and glutamine

Second-order Møller-Plesset perturbation (MP2) theory, Hartree-Fock (HF) theory, and 25 density functional theory (DFT) methods have been used to compute the hydrogen bond strengths between water and the polar side chains of four uncharged amino acids (serine, threonine, asparagine, and glutamine) and their corresponding glycine-X-glycine tripeptides (where X is one of the four amino acids just listed). These theoretical methods were also combined in 26 different integrated ONIOM QM:QM schemes in which MP2 was used as the high level method and either HF theory or one of the 25 density functionals was used as the low level method. Three important observations lead to the conclusion that the MP2:HF hybrid method provides the best description of hydrogen bonding in these complexes. (1) With deviations from the MP2 dissociation energies that range from −0.42 to −1.06 kcal mol−1 and an average absolute error (AAE) of 0.69 kcal mol−1, MPWLYP is the most accurate density functional. Only the MPWLYP, B971, X3LYP, PBE1PBE, and VSXC functionals have AAEs under 1 kcal mol−1 while the maximum absolute error exceeds 1 kcal mol−1 for all 25 functionals. (2) The integrated ONIOM methods outperform the DFT methods, including the MPWLYP functional. The ONIOM AAEs range from 0.25 to 0.36 kcal mol−1, all of which are significantly smaller than the corresponding value for the MPWLYP functional. (3) None of the MP2:DFT approaches is better than the MP2:HF scheme. MP2:HF has the smallest AAE (0.25 kcal mol−1), the narrowest range of errors (−0.35 to −0.12 kcal mol−1), and the smallest maximum absolute error (0.35 kcal mol−1) of all 26 schemes.