Evaluation of a combined energy fitness function for a distributed memetic algorithm to tackle the 3D protein structure prediction problem

This paper addresses the problem of predicting the tertiary structure of a given amino acid sequence, which has been reported as a NP-Complete class problem. We evaluate how the combination of two weighted energy functions, Talaris and SASA from Rosetta software suite, can improve, in reasonable computational time, the quality of the solutions in a distributed implementation of a knowledge-based memetic algorithm for the protein structure prediction problem. The proposal is evaluated in terms of prediction accuracy and biological significance of the predicted structures. Results reveal that the packing characteristic of the SASA function accelerates the search of good solutions in terms of root-mean-square deviation while a small weighting of the Talaris function helps to maintain the stability of their internal and interactions free energies, allowing the predictions of structures comparable to the experimental results.