Protein structure prediction using a hybrid energy function and an exact enumeration

Abstract

We develop a protein structure prediction method that utilizes fragment assembly and a hybrid energy function. In a fragment assembly method, the local structure of the backbone is obtained from a structural database by using similarity of sequence features, in contrast to a pure physics-based method in which all dihedral angles are allowed to vary continuously. Since the conformational space for the backbone is finite, we generate all possible conformations and vary only the side-chain dihedral angles for each of them. The conformations are scored using a hybrid energy function, where all the backbone atoms are described explicitly, but the side chain is modeled as a few interaction centers. We perform a test prediction on four proteins, 112y, 1e01, 1bdd and 1bk2, to demonstrate the feasibility of protein structure prediction based on exact enumeration.