The Influence of Mutation on Protein-Ligand Docking Optimization: A Locality Analysis

Evolutionary approaches to protein-ligand docking typically use a real-value encoding and mutation operators based on Gaussian and Cauchy distributions. The choice of mutation is important for an efficient algorithm for this problem. We investigate the effect of mutation operators by locality analysis. High locality means that small variations in the genotype imply small variations in the phenotype. Results show that Gaussian-based operators have stronger locality than Cauchy-based ones, especially if an annealing scheme is used to control the variance.