Assessment of RNA Force Fields for Dynamic Docking of Small Molecules Using Multicanonical MD Simulations

Molecular dynamics (MD) simulations are increasingly important for analyzing RNA-ligand interactions, particularly in the context of therapeutic development. However, the accuracy of RNA force fields remains insufficiently assessed, partly due to the limited sampling efficiency of MD approaches and the lack of reliable docking protocols. To evaluate the performance of modern AMBER-based RNA force fields, we selected four small RNA-ligand complexes from the Protein Data Bank (PDB) and executed dynamic docking simulations using one of the generalized ensemble methods, multicanonical MD, across five different force fields. We analyzed a total of 600 μs of simulation data, each reweighted to the canonical ensemble at physiological temperature. The resulting conformational ensembles varied across force fields for three of the four targets. Among the tested force fields, the parm99χOL3-vdW_bb yielded the most accurate results based on our R-value analysis that measures the ligand-RNA native contacts, assuming the PDB structures represent the correct native conformations. However, further analysis revealed that some metastable, non-native RNA conformations had smaller intercalation sites with a closed binding pocket, resulting in shrinkage of the RNA molecules. These findings suggest that current RNA force fields may overstabilize non-native, closed conformations. The present simulations, methodology, analyses, and data offer valuable insights to guide the development of next-generation RNA force fields to better assess non-native RNA conformations.