Refinement of molecular dynamics ensembles using experimental data and flexible forward models

Thorben Fr\"ohlkingMattia BernettiGiovanni Bussi

Mar 2023

摘要

A novel method combining maximum entropy principle, the Bayesian-inference ofensembles approach, and the optimization of empirical forward models ispresented. Here we focus on the Karplus parameters for RNA systems, whichrelate the dihedral angles of $\gamma$, $\beta$, and the dihedrals in the sugarring to the corresponding $^3J$-coupling signal between coupling protons.Extensive molecular simulations are performed on a set of RNA tetramers andhexamers and combined with available nucleic-magnetic-resonance data. Withinthe new framework, the sampled structural dynamics can be reweighted to matchexperimental data while the error arising from inaccuracies in the forwardmodels can be corrected simultaneously and consequently does not leak into thereweighted ensemble. Carefully crafted cross-validation procedure andregularization terms enable obtaining transferable Karplus parameters. Ourapproach identifies the optimal regularization strength and new sets of Karplusparameters balancing good agreement between simulations and experiments withminimal changes to the original ensemble.

展开全部

图表提取