Performance of empirical potentials (AMBER, CFF95, CVFF, CHARMM, OPLS, POLTEV), semiempirical quantum chemical methods (AM1, MNDO/M, PM3), andab initio Hartree-Fock method for interaction of DNA bases: Comparison with nonempirical beyond Hartree-Fock results

Abstract

Empirical energy functions (AMBER 4.1, CFF95, CHARMM23, OPLS, Poltev), semiempirical quantum chemical methods (AM1, MNDO/M, PM3), and the nonempirical ab initio self-consistent field (SCF) method utilizing a minimal basis set combined with the London dispersion energy (SCFD method) were used for calculation of stabilization energies of 26 H-bonded DNA base pairs, 10 stacked DNA base pairs (thymine was replaced by uracil), and the B-DNA decamer (only DNA bases were considered). These energies were compared with nonempirical ab initio beyond Hartree–Fock values [second-order Moller–Plesset (MP2)/6–31G*(0.25)]. The best performance was exhibited by AMBER 4.1 with the force field of Cornell et al. The SCFD method, tested for H-bonded pairs only, exhibited stabilization energies that were too large. Semiempirical quantum chemical methods gave poor agreement with MP2 values in the H-bonded systems and failed completely for stacked pairs. A similar failure was recently reported for density functional theory calculations on base stacking. It may be concluded that currently available force fields provide much better descriptions of interactions of nucleic acid bases than the semiempirical methods and low-level ab initio treatment. © 1997 John Wiley & Sons, Inc. J Comput Chem18: 1136–1150