Selection of single nucleotide polymorphisms and genotype quality for genomic prediction of genetic merit in dairy cattle
G. R. Wiggans*1, T. S. Sonstegard1, P. M. VanRaden1, L. K. Matukumalli1,2, R. D. Schnabel3, J. F. Taylor3, F. S. Schenkel4, C. P. Van Tassell1; ARS, USDA, Beltsville, MD, USA1, George Mason University, Manassas, VA, USA2, University of Missouri, Columbia, MO, USA3, University of Guelph, Guelph, ON, Canada4
A process to prepare high-density genotypic data for use in genomic prediction was developed. Marker genotypes from >51,000 single nucleotide polymorphisms (SNP) were generated for 3,139 Holstein bulls on the Illumina Bovine SNP50 chip. The SNP were grouped by minor allele frequency (MAF); 10,249 SNP with a MAF of <5% were excluded. Number of SNP for each of 45 MAF categories was uniform (800 to 1,009). Hardy-Weinberg equilibrium was assessed by comparing observed to expected heterozygosity for each locus. For 6 SNP assigned to chromosome 7, no bulls were heterozygous, which confirms the latest assembly that places those SNP on the X chromosome. Observed heterozygosity was within 2% of that expected for 96% of SNP. Linkage between adjacent autosomal SNP was analyzed to determine if the data set could be reduced for downstream analysis. For 1,237 pairs of adjacent SNP, marker genotypes were either both homozygous or both heterozygous (<10 bulls differed for each pair), and the first SNP from each pair was excluded; mean physical distance between those SNP pairs was much smaller (37 kb) than between 39,386 autosomal SNP (64 kb). Sire and son data for 2,566 bulls with 204 genotyped sires were compared to validate sample identification and Mendelian inheritance. For those bulls with >100 inheritance errors, correct sire was determined through comparison with other sires of sons. For sons with the correct sire, 99.99% of SNP with genotypes did not conflict. Comparison of genomic and pedigree relationships detected 3 members of a clonal family, a set of identical twins, and some possible pedigree errors. Genotyping consistency was investigated for 9 bulls genotyped twice and for the twins and clones. Most differences were caused by an inability to determine the genotype for one of the paired SNP; however, one clone had 24 SNP conflicts (99.94% concordance). Although evaluation of the SNP set is ongoing, only minor changes are expected for the final set. This largest set of high-quality SNP data for Holsteins to date should provide the basis for successful genomic prediction.
KEYWORDS
genomic prediction
genotyping
single nucleotide polymorphism
|
