Here we present, HaploPool, a novel and cost-effective method for estimating haplotype frequencies from pooled DNA samples. We assume that the genotyping is done on many unrelated diploid individuals which are pooled into disjoint pools of a small number of individuals (usually two or three individuals per pool). HaploPool is an implementation of two novel and complementary algorithms: one based on a model of the perfect phylogeny haplotyping problem and the other based on a least squares regression model of linear haplotype frequency constraints.
We compared HaploPool to three programs for haplotype frequency estimation from pool genotypes. For an objective standard, we also compared HaploPool to the state-of-the-art haplotype frequency estimation program for non-pool genotypes. HaploPool runs considerably faster (at least six times faster) than any of the four other programs. This means that it is feasible to estimate haplotype frequencies for the whole genome, after partitioning the genome into blocks of (5-25 SNPs). In addition, pooled DNA is a cost-effective strategy when compared to non-pooled DNA. Assuming that genotyping is more expensive than the sample collection and phenotyping procedures, our results show that pool genotyping and haplotype frequency estimation with HaploPool is more cost-effective than non-pool genotyping and frequency estimation. Because these assumptions often hold for species grown in laboratories, HaploPool would be useful for conducting cost-effective association studies both on model organisms and on organisms about which little is known.