Comparing the performance of a knowledge-based potential and a knowledge-based force functions in the scoring of protein-protein complexes

The knowledge-based force function is a new type of the scoring functions that has been used in the field of protein fold recognition with a noticeable success. In this study we compared the performance of a knowledge-based potential function and its corresponding force function in discrimination of the correct protein-protein complexes from the incorrect ones. The total force imposed by one component (receptor/ligand) upon another was used as a measure of complex stability. This force is expected to be the lowest in the native structure. To test the performance of each method, two decoy sets were used; one generated by soft body docking and the other by rigid body docking algorithms. The results of this comparison show that, for both decoy sets, the success rates in native and near native selections of the energy model are higher than that of the force model. It seems, the dependence of amount of the force on shape of the interface region introduces errors in the later model and therefore makes it inappropriate for the scoring of docked complexes.