110 Refinement of force field torsion parameters for nucleic acids based on inclusion of conformation-dependent solvation effects

Abstract

Accurate representation of nucleic acids in molecular dynamics simulations depends critically on the quality of the applied empirical force field. Among force field terms, the torsion parameters are known to strongly influence the conformational equilibria and molecular structures. Unfortunately, past several years witnessed severe problems in describing the torsion space in nucleic acids by current force fields and more problems continue emerging. In an attempt for improvement, we suggested a novel parameterization procedure that incorporates some previously neglected solvation-related effects, which proved to be essential for obtaining accurate torsion profiles. The suggested approach avoids double counting of solvation effects and provides parameters that may be used in combination with any of the widely used nonpolarizable discrete solvent models or with the continuum solvent models. Improvements are demonstrated for the latest AMBER force field for RNA simulations, ff10, which incorporates parameters for the glycosidic torsion (χOL3) developed by us using the above-described procedure (Banáš et al., 2010; Zgarbová et al., 2011). Resulting parameters are verified by extensive molecular dynamics simulations of canonical RNA duplexes and RNA hairpin loops. We show that our modification removes overstabilization of the high-anti region found in the ff99 force field and thus prevents formation of undesirable ‘ladder-like’ structural distortions in RNA simulations. In addition, we applied our parameterization approach to development of the glycosidic torsion in DNA (χOL4). This refinement focuses on adjusting description of the syn region and syn-anti balance of the χ potential. This modification exhibits a notable improvement of the description of the antiparallel G-DNA stem, which was not modeled correctly by the current ff99 force field (Krepl et al., 2012).