Submitted to: Journal of Dairy Science
Publication Type: Abstract Only
Publication Acceptance Date: 3/8/2010
Publication Date: 6/24/2010
Citation: Van Raden, P.M., Tooker, M.E. 2010. Gains in reliability from combining subsets of 500, 5,000, 50,000 or 500,000 genetic markers. Journal of Dairy Science. 93(E-Suppl. 1):534(abstr. 620).
Technical Abstract: More genetic markers can increase both reliability and cost of genomic selection. Fewer markers can be used to trace chromosome segments within a population once identified by high-density haplotyping. Combinations of marker densities can improve reliability at lower cost. As of January 2010, 33,414 North American Holsteins had been genotyped for 50,000 genetic markers. Genotypes for 500,000 markers were simulated using pedigree data for this same population. Linkage was introduced among base alleles to make correlations among simulated genotypes similar to actual. Reduced subsets were examined using every 10th, 100th, or 1000th marker. In marker regression models, polygenic variance was 70, 30, 10, and 0% of genetic variance with 500, 5,000, 50,000 and 500,000 markers, respectively. Respective reliabilities obtained as squared correlations of estimated and true breeding values averaged across 5 replicates were 39.4, 70.2, 82.6, and 84.0% for 14,061 young bull predictions. At highest density, one processor required 2.5 d to complete 150 iterations for the 5 replicates. A mixed-density data set had 500,000 markers genotyped for 3,515 young bulls and 3,883 bulls with >90% reliability and 50,000 markers genotyped for the remaining 26,016 animals. This dataset had 70% missing genotypes; however, after imputing from haplotypes, only 4% of genotypes were missing, and average reliability was 83.1%. Two other mixed-density data sets had 50,000 markers for cows and progeny-tested bulls but only 5,000 or 500 markers for young animals. Reliabilities averaged 79.6% for young animals if 5,000 markers were genotyped and the other 45,000 imputed. At 500-marker density, inheritance probability was computed for each marker instead of simply assigning either parental haplotype; reliabilities averaged 70.3% when young animals were genotyped for 500 markers and both parents were genotyped for 50,000. Very high marker density can increase reliability slightly (1.4%), whereas low marker density allows breeders to apply cost-effective genomic selection to many more animals.