Developing Explicit Base Stacking Potentials for the IsRNA2+ Coarse-Grained RNA Force Field Using Iterative Reweighting.

Accurate prediction of RNA tertiary structures is essential for understanding their diverse biological functions. While machine learning methods have gained popularity, physics-based coarse-grained force fields remain an important tool for elucidating the structure and dynamics of RNA molecules. In this paper, we introduce the iterative reweighting algorithm as a flexible framework for optimization of force field parameters. We apply this framework to develop the IsRNA2+ force field, where the IsRNA force field is augmented by an explicit base-stacking potential. The performance of the IsRNA2+ force field is validated through structure prediction tasks on a benchmark set of 46 "simple" and 20 "difficult" RNAs. Compared to the original IsRNA2 force field, the current IsRNA2+ force field leads to improved accuracy in structure prediction. A detailed analysis of several cases with large improvements highlighted the importance of a balanced treatment of local energy components in determining the overall tertiary structures of RNA molecules. We also envision the potential of combining the machine learning approach with traditional force fields to further improve the understanding of RNA structures and dynamics.