(362b) Tackling Force-Field Bias in Protein Folding Simulations

Computer simulations can potentially provide molecular details of protein folding mechanism at time- and length-scales which are not easily accessible in a laboratory experiment. Advances in computational resources have allowed the "brute force" calculation of folding trajectories with all-atom details in explicit solvent. These have highlighted an under appreciated shortcoming of energy functions used in these simulations, i.e., most have significant biases in secondary structure propensity. This problem of inherent bias in current energy functions usually requires the energy function to be chosen according to the folded structure of the protein (i.e., helical versus beta-sheet). However, the ultimate aim is to find a ?transferable? force field which is only dependent on protein sequence and not on the final native structure.

We show that a simple correction to the backbone potential results in an unbiased energy function for protein folding as judged by several protein models, e.g., GB1 hairpin, Trp cage, Villin, Pin WW domain. Starting from unfolded configurations, we obtain folding events for these proteins within 2 Å RMSD of the experimental structures. We obtain converged equilibrium distributions, with folded populations at standard conditions (300 K) in remarkable accord with experiment. Comparison to experimental data from NMR spectroscopy and FRET suggests that while the folded structures are accurately reproduced, the unfolded state remains too structured and compact.