Computational experiments on algorithms for haplotype inference problems by pure parsimony

To analyze the function of DNA, researchers have to obtain each haplotype, the genetic constitution of an individual chromosome, of an individual for analysis. Due to the significant efforts required in collecting haplotypes, the descriptions of one conflated pair of haplotypes called genotypes are usually collected. Since the genotype data contains insufficient information to identify the combination of DNA sequence in each copy of a chromosome, one has to solve the population haplotype inference problem by pure parsimony criterion which uses the minimum number of haplotypes to infer the haplotype data from genotype data for a population. Previous researches use mathematical programming methods such as integer programming and semidefinite programming models to solve the population haplotype inference problem. However, no computational experiment has ever been conducted to evaluate the algorithmic effectiveness. This paper thus conducts the first computational experiments on four haplotyping algorithms, including our new greedy heuristic and three pervious haplotyping algorithms.