BICePs repx: Force field validation by a Bayesian/maximum entropy approach.

Abstract

Bayesian Inference of Conformational Populations (BICePs) is a reweighting algorithm that reconciles simulated ensembles with sparse and/or noisy observables. Typically, force fields are ranked using a chi-squared metric by comparing simulated observables with experimental measurements, using some estimate of the experimental uncertainty, while neglecting systematic error. In contrast to this approach, BICePs has the significant advantage of performing objective model selection by sampling the full posterior distribution of conformational populations as well as uncertainties due to random and systematic error. To demonstrate this, we used BICePs to perform Bayesian model selection for nine different force fields used to simulate the folding dynamics of the mini-protein chignolin in TIP3P solvent. The simulated ensembles were compared against 158 experimental measurements (139 NOE distances, 13 chemical shifts, and 6 vicinal J-coupling constants for HN and Hα). A significant advantage of our approach is the BICePs score, which is a metric that quantifies the agreement between simulation and experiment by computing a free energy-like quantity that reflects the amount of adjustment/reweighting necessary for the populations to match the experimental restraints. For the nine force fields tested (ff14SB, ff99SB-ildn, ff99, ff99SBnmr1-ildn, ff99SB, CHARMM22star, CHARMM27, CHARMM36, OPLS-aa), our results are consistent with previous work that used a conventional chi-squared metric for model selection for small polypeptides and ubiquitin, with ff99SBnmr1-ildn remaining as the best-scoring model.