A Physics-Based Iterative Method to Extract Distance-Dependent All-Atom Potentials for Protein Structure Prediction

One of the challenges in protein structure prediction is the development of an accurate scoring function that yields a global minimum free energy for the native state. Although the knowledge-based scoring function has proven to be a successful scoring approach to protein structure prediction, there exists a hurdle in deriving knowledge-based potentials. Namely, the ideal reference state is inaccessible. In this work, we have developed a general physics-based iterative method to extract distance-dependent all-atom potentials for protein structure prediction. Our method circumvented the long-standing reference state problem. The derived scoring function was extensively evaluated with three diverse test sets, and showed significant improvement over other well-known scoring functions. The results suggest the efficacy of our scoring function for protein structure prediction.