Comparison of phasing strategies for whole human genomes

• DOI: 10.1371/journal.pgen.1007308
• Publication Date: 2018-04-06T00:09:38Z
• Authors: Yongwook Choi; Agnes P. Chan; Ewen Kirkness; Amalio Telenti; Nicholas J. Schork
• Keywords: Phaser

Humans are a diploid species that inherit one set of chromosomes paternally and homologous maternally. Unfortunately, most human sequencing initiatives ignore this fact in they do not directly delineate the nucleotide content maternal paternal copies 23 individuals possess (i.e., 'phase' genome) often because costs complexities doing so. We compared 11 different widely-used approaches to phasing genomes using publicly available 'Genome-In-A-Bottle' (GIAB) phased version NA12878 genome as gold standard. The strategies we included laboratory-based assays prepare DNA unique ways facilitate well purely computational seek reconstruct phase information from general reads constructs or population-level haplotype frequency obtained through reference panel haplotypes. To assess performance approaches, used metrics included, among others, switch error rates, block lengths, proportion fully phase-resolved genes, accuracy yield between pairs SNVs. Our comparisons suggest hybrid combined approach leverages: 1. population-based SHAPEIT software suite, 2. either genome-wide read data parental genotypes, 3. large variant frequencies, provides fast efficient way produce highly accurate individual genomes. found for is enhanced with addition data; e.g., whole shotgun and/or RNA reads. Further, inclusion genotype within strategy can provide much ten-fold reduction errors. also considered majority voting scheme construction consensus combining multiple predictions site coverage. Finally, identified sequence signatures associated genomic regions harboring errors, which low polymorphism SNV density.