CRANBERRY: Coarse-grained Model of RNA Dynamics and Thermodynamics Allowing for Sugar Puckering and Noncanonical Base Pairing

RNA dynamics play a crucial role in cellular functions. Computational models provide molecular-level insights for RNA dynamics. However, current all-atom RNA force fields suffer from energy imbalance issues, such as over-stabilizing stacking interactions by 1-2 kcal/mol and causing disordered RNA to collapse. Additionally, these models are computationally inefficient for studying biologically relevant dynamics. While coarse-grained models address the efficiency limitation, they often fail to account for realistic non-canonical base pairing and sugar puckering. To overcome these challenges, we developed CRANBERRY, a coarse-grained model which incorporates these features. The force field was trained via our contrastive divergence method using multiple NMR ensembles. CRANBERRY demonstrates native stability comparable to all-atom force fields and outperforms RACER, another coarse-grained model aimed at capturing both structures and thermodynamics. Moreover, CRANBERRY more accurately reproduces experimental stacking free energies and prevents the collapse of disordered RNA. CRANBERRY allows for computationally efficient and accurate modeling of complex RNA dynamics.