Ensemble-Based Precision Refinement of All-Atom Nucleic Acid Force Fields Guided by NMR NOE Pair-Distance Measurements.

Accurately modeling nucleic acid structure and dynamics remains challenging for all-atom simulations, especially for noncanonical motifs such as small loops and G-quadruplexes. Despite these advances, current all-atom classical force fields often fail to reproduce ensembles consistent with high-resolution experimental data. We present a systematic refinement strategy for AMBER-based force fields that incorporates nuclear Overhauser effect distance data from NMR experiments within an ensemble-averaged optimization framework. By selectively tuning van der Waals interaction pairs, this approach markedly reduces simulation-experiment discrepancies, removes persistent artifacts, and generates free energy landscapes that better reflect experimental observations. We demonstrate broad applicability across diverse DNA and RNA systems including flexible loops and G-quadruplexes. Overall, this transferable strategy significantly improves structural accuracy and predictive power, enabling more reliable modeling of complex nucleic acid conformational ensembles.