The Short and the Long of It: A Comparison of Short and Long Haplotypes in Dairy Cattle

Low (19K, 19,000 SNP), medium (50K; 54,609 SNP) and high-density (HD; 777,962 SNP) SNP arrays, and more recently whole genome sequence (WGS) data, are revolutionizing animal breeding. In dairy cattle populations, the population structure comprising large half-sib families facilitates imputation of animal genotypes from array-based densities up to sequence level using WGS information of key ancestors. Genotypes can be phased to reconstruct the haplotypes that represent the original allelic combinations that an individual received from its parents, equating to in silico reconstruction of parental haplotypes.

Phasing and imputation methods are often broadly divided into family-based methods, which use linkage information from close relatives, and population-based methods, which use population linkage disequilibrium information. Family-based methods usually use long haplotypes, where population-based methods use shorter haplotypes in terms of number of SNPs and/or haplotype length in Mb. Shorter haplotypes could be more useful for across-breed analyses (e.g. genomic selection or GWAS for historical mutations), however require a much larger sample size for higher accuracy. Genotyping errors and especially misplaced SNPs may disturb long range phasing (observed more in higher density panels). Longer haplotypes are likely helpful to detect recent mutations (they appear only on long haplotypes, and have very low frequency).

The correlations of SNP alleles within linkage disequilibrium (LD) blocks (local or short-range phasing) is related to haplotype length. In this study we compared both 50K and HD haplotypes determined using short and long windows. Window sizes used for 50K data were: 11, 20, 50, 100, 200, 500, 1000 SNPs / window. For HD data, window sizes were: 100, 200, 500, 1000, 3200, 8200, 16800 SNPs / window, covering 0.6, 1.2, 2.9, 5.9, 11.8, 29.5 and 59.0 Mb for 50K and HD, respectively.

The average number of heterozygotes and the average relative haplotype accuracy (# of incompatible haplotypes / # heterozygotes) was evaluated for Brown Swiss, Braunvieh and Original Braunvieh animals. Longer haplotypes were generally more accurate than short haplotypes, however accuracy depended on relationships between individuals and likely on assembly accuracy.