Diversity control in genetic algorithms for protein structure prediction

In recent years, there is a growing interest in using Genetic Algorithms (GAs) in the protein structure prediction problem. However, the search space in this problem is very complex, what results in premature convergence of the GAs in their standard form, as the population generally gets trapped into local optima. Based on this fact, the use of two different strategies that can help GAs to maintain or increase the diversity of the population in the protein structure prediction problem are investigated in this paper. These strategies are Hypermutation and Random Immigrants. A new form of codification of the protein structure in the GA using sorted angles database is still proposed. Experimental results with Crambin (PDB code 1CRN), Met-Enkephalin (PDB code 1PLW), and DNA-Ligand (PDB code 1ENH) show that strategies to increase or maintain the population diversity are interesting for the protein structure prediction problem.

Publication year:2009

Accessibility:Open