WinHAP2: an extremely fast haplotype phasing program for long genotype sequences

Abstract

BackgroundThe haplotype phasing problem tries to screen for phenotype associated genomic variations from millions of candidate data. Most of the current computer programs handle this problem with high requirements of computing power and memory. By replacing the computation-intensive step of constructing the maximum spanning tree with a heuristics of estimated initial haplotype, we released the WinHAP algorithm version 1.0, which outperforms the other algorithms in terms of both running speed and overall accuracy.ResultsThis work further speeds up the WinHAP algorithm to version 2.0 (WinHAP2) by utilizing the divide-and-conquer strategy and the OpenMP parallel computing mode. WinHAP2 can phase 500 genotypes with 1,000,000 SNPs using just 12.8 MB in memory and 2.5 hours on a personal computer, whereas the other programs require unacceptable memory or running times. The parallel running mode further improves WinHAP2's running speed with several orders of magnitudes, compared with the other programs, including Beagle, SHAPEIT2 and 2SNP.ConclusionsWinHAP2 is an extremely fast haplotype phasing program which can handle a large-scale genotyping study with any number of SNPs in the current literature and at least in the near future.

Unzipping haplotypes in diploid and polyploid genomes

“…Currently, most state-of-the-art algorithms employ the alignment-based strategy to determine adjacent loci based on Illumina short reads and/or PacBio long reads [8], such as HAPCUT2 [9], WHATSHAP [14] and WinHAP2 [41]. However, phasing with Illumina short reads often produce limited length of phased blocks since the distance of adjacent polymorphic sites can exceed the length of a typical Illumina read or read pair.…”

Mapping Phylogenetic Trees to Reveal Distinct Patterns of Evolution

“…Sequencing technologies continue to decrease in cost, with the result that it is now feasible to sequence up to tens of thousands of taxa in multiple genes, at least in viruses. As the computational challenges associated with haplotype phasing are resolved ( Pan et al 2014 ; Zhi and Zhang 2014 ; Regan et al 2015 ), phylogenetic methods will be used for large data sets on higher organisms. As an example, the Epidemiology Network Ag1000G has 765 Anopheles mosquito genomes visible to the public ( MalariaGEN 2015 ).…”