An Information Based Genetic Algorithm Approach to Fast Peptide Docking

In genetic algorithm (GA), there are two main methods of determining trial candidates: crossover and mutation. While crossover directs search between fit candidates, mutation plays a role on jumping out local optimal. In molecular docking calculations, it is desirable to chart as much unexplored search space as possible. Therefore it is desirable that mutation results in a uniform distribution of sampling points in solution space. In this work an information entropy based mutation procedure is developed. Instead of random mutation, mutation is directed to parts of the solution space that is least populated. Such a procedure is implemented in AUTODOCK and used to study the docking of Peroxisome Proliferator-Activated Receptors gamma (PPARResults show that the proposed information-based genetic algorithm is superior to conventional GA for docking both in speed and precision.