Impact of Pedigree Information and Genome Assembly Errors on Inference of SNP Haplotypes in Cattle

The accuracy of genotype imputation depends on the accuracy of phasing haplotypes from single nucleotide polymorphism (SNP) genotypes. One approach to judge phasing accuracy is to estimate the number of recombination events by comparing diplotypes within parent/offspring pairs. The study evaluated the impact of pedigree information and SNP location determined from genome map on phasing accuracy, by quantifying recombination events. Among BovineSNP50k genotypes on 2,778 parent verified Angus sire/offspring pairs, SNPs were removed with call rate <0.95, minor allele frequency <0.01, p value of a Hardy Weinberg Equilibrium test <0.001, and Mendelian inconsistency rate >0.0024 (95% quantile). Phasing of Bos taurus (BTA) chromosomes 14 and 15 were performed using the unrelated option of BEAGLE (Browning and Browning, 2007) or the pedigree-based DAGPHASE (Druet and Georges, 2010) each with two genome maps UMD3.1 or USDA-Animal Improvement Programs Laboratory (AIPL) map. The expected numbers of recombination events were estimated using the Haldane map function and for BTA14 were (#cattlex#recombination) 1187x0, 1009x1, 429x2, and 153x>2; and for BTA15 were 1176x0, 1011x1, 435x2, and 156x>2. Average numbers of cattle observed with >2 recombination events were 928 using BEAGLE and 467 using DAGPHASE. Both methods overestimated recombination events, but DAGPHASE, using a single generation pedigree was superior to BEAGLE. The average number of cattle with >2 recombination events estimated using DAGPHASE and UMD3.1 locus ordering was 544, but was 417 using the USDA-AIPL map. Recombination hotspots were detected near 4 and 82Mb on BTA14, and near 25Mb on BTA15.