Fragment based protein structure prediction. (c2013)

Abstract

In recent years, the protein structure prediction problem has come under extensive investigation, and computational structure prediction methods are circumventing the time-consuming experimental methods by accelerating the prediction process. This work presents a fragment based protein tertiary structure prediction method that provides suboptimal structures. In addition, it demonstrates the advantage of using the CHARMM36 energy model. The method is based on a two-phase Scatter Search metaheuristic that minimizes the energy function. Backbone fragments selections extracted from the Robetta server are followed by side chain selections, extracted from the Dunbrack Library. The results, evaluated on three proteins, show that the algorithm produces tertiary structures with promising root mean square deviations, within reasonable times.