A Multiobjective Algorithm for Protein Structure Prediction Using Adaptive Differential Evolution

Protein Structure Prediction (PSP) is one of the most challenging problems in Bioinformatics research area. This paper models PSP as a multiobjective optimization problem and adopts Adaptive Differential Evolution for Multiobjective Problems (ADEMO/D) to minimize potential energies (bonded and non-bonded) providing final protein structures. ADEMO/D incorporates concepts of Multiobjective Evolutionary Algorithms based on Decomposition (MOEA/D) and mechanisms of mutation strategies adaptation. In this work the probability matching and extreme absolute reward methods are combined to adapt ADEMO/D to the PSP context. The DE mutation strategy is chosen from a candidate pool according to a probability that depends on its received reward. We test the behavior of the proposed method, considering the off-lattice model and ab initio approach for PSP, in Met-Enkephalin peptide and 1ZDD protein. The results point ADEMO/D as a competitive approach for potential energy values and conformation similarity metrics.