Protein Structure Prediction with Stochastic Optimization Methods: Folding and Misfolding the Villin Headpiece

Abstract

AbstractWe recently developed an all-atom free energy forcefield (PFF01) for protein structure prediction with stochastic optimization methods. Using this forcefield we were able to reproducibly fold the 20 amino-acid trp-cage protein and the 40-amino acid three-helix HIV accessory protein. We could also demonstrate that PFF01 stabilized the native folds of various other proteins ranging from 40-60 amino acids at the all atom level. Here we report on a folding study on the widely investigated autonomously folding 36-amino acid villin headpiece. Using more than 76000 low-energy decoys to characterize its free-energy landscape, we find several competing low-lying three-helix structures. The existence of these metastable conformations, which are not nearly as prevalent in other proteins, may explain the extreme difficulty in folding this protein in-silico.KeywordsPotential Energy SurfaceFree Energy SurfaceTerminal BranchProtein Structure PredictionSecondary Structure ContentThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.