Haplotype inference using a genetic algorithm

The haplotype inference problem is a computational task to infer haplotype pairs based on the phaseunknown genotypes, and is pivotal in the International Hapmap project. The haplotype inference problem is NP-hard, and exact algorithms become infeasible when the problem sizes are big. Genetic algorithms (GA) are commonly used to approximate optimal solutions for NP-hard problems within reasonable computation time. In this paper, we have proposed a simple genetic algorithm formulation for the haplotype inference problem based on the model of parsimony, which aims to resolve the existing genotypes using as few haplotypes as possible. We applied our GA in the real datasets of the human beta₂AR locus and APOE locus, and compared the solutions to the experimentally verified haplotypes; we have found that our approach of inferring haplotypes is very accurate. We believe that our GA is a potentially powerful method for robust haplotype inferences.