Inconsistencies in ab initio evaluations of non-additive contributions of DNA stacking energies

TL;DR

This study investigates the non-additive inter-molecular interactions in DNA stacking energies using diffusion Monte Carlo methods, revealing significant discrepancies with other ab initio approaches and raising questions about methodological accuracy.

Contribution

It highlights inconsistencies in non-additive energy evaluations of DNA stacking by FNDMC compared to other ab initio methods, emphasizing the need to understand fixed node biases.

Findings

FNDMC results differ significantly from other ab initio methods for non-additive contributions.

Stacking energies show no remarkable differences across methods.

Error cancellations in FNDMC may not apply to non-additivity involving charge transfer.

Abstract

We evaluated the non-additive contributions of the inter-molecular interactions in B-DNA stacking by using diffusion Monte Carlo methods with fixed node approximations (FNDMC). For some base-pair steps, we found that their non-additive contributions evaluated by FNDMC significantly differ from those by any other {\it ab initio} methods, while there are no remarkable findings on their stacking energies themselves. The apparently unexpected results of non-additivity raise issues in both FNDMC and correlated wavefunction methods. For the latter, it can be partly attributed to the imperfect complete basis set (CBS) correction scheme due to the limitation of the computational costs. On the other hand, the striking contrast between the stacking and non-additivity behaviors was found in FNDMC. This might imply that the error cancellations of the fixed node biases in FNDMC work well for the stacking energies, while not for the non-additivity contributions involving charge transfers caused by hydrogen bonds bridging Watson-Crick base pairs.