A Generalized Framework for Developing Dihedral Torsion Energy Terms in Nucleic Acids Force Fields

Chungwen Liang, David Pekker, Alessio Valentini, Swagatam Mukhopadhyay

Abstract: Accurate modeling of large-scale biomolecular systems depends on high-quality force field (FF) models. While atomistic nucleic acid FFs have improved over the past four decades, a universal framework for both natural and chemically modified nucleic acids across diverse environments remains lacking. In this work, we introduce a general methodology for developing torsional energy parameters that apply to all nucleic acid systems. Our approach simultaneously parametrizes key dihedral angles in nucleic acids critical to simulating their conformations at physiologically relevant temperatures and solvent environments. The resulting FF, Creyon25, achieves accuracy comparable to the latest AMBER and CHARMM models, but our framework in contrast is generalizable to chemical modifications (in linker, sugar and base). We validate Creyon25 across a wide range of RNA and DNA structures, including tetramers, tetraloops, and duplexes. We found that the Creyon25 RNA model accurately reproduces experimentally observed structures, although there is still room for improvement in the Creyon25 DNA model. This work represents a major step toward creating robust FFs for chemically modified nucleic acids, supporting the advancement of oligonucleotide therapies.