Investigations of Stacked DNA Base-Pair Steps: Highly Accurate Stacking Interaction Energies, Energy Decomposition, and Many-Body Stacking Effects.

Abstract

The stacking energies of 10 unique B-DNA base-pair steps were calculated with highly accurate quantum chemistry and used as reference values in a thorough benchmark of (dispersion-corrected) DFT, wave function methods, tight-binding methods, and different force fields, including charge variants thereof. The reference values were computed using a focal-point energy function based on extrapolated explicitly correlated MP2-F12 and conventional CCSD(T) data at the triple-ζ level. A collection of 29 different density functionals, sometimes with multiple dispersion corrections (D3(BJ), D3M(BJ), and VV10) were evaluated, including recent functionals like B97M-V, ωB97M-V, and SCAN-D3(BJ), which perform excellently. The double-hybrid DSD-BLYP-NL was found to be the best DFT method. Common wave function methods (MP2, SCS-MP2, and MP2.5) and the SNS-MP2 protocol were tested as well, where only the latter and DLPNO-CCSD(T)/CBS were competitive with DFT. The tight-binding methods DFTB3-D3 and GFN-xTB revealed a comparatively low accuracy. The AMBER force field outperformed CHARMM and GROMOS but still showed systematic gas-phase overbinding, which could be traced back to the electrostatic term, as revealed by comparison of different sets of point charges. High-order SAPT, e.g., SAPT2 + 3δ(MP2), was not only benchmarked but also used to study the nature of the stacking interactions to high accuracy. The δ(MP2) term turned out to be crucially important to reach high accuracy. Finally, we investigated four-body stacking effects with DLPNO-CCSD(T) and DFT, which were found to be significant and strongest for the CpC base-pair step where they reached almost 30% of the total stacking energy.