Submitted to: Interbull Annual Meeting Proceedings
Publication Type: Proceedings
Publication Acceptance Date: 5/30/2012
Publication Date: 6/1/2012
Citation: Van Raden, P.M., Olson, K.M., Null, D.J., Sargolzaei, M., Winters, M., Van Kaam, J. 2012. Reliability increases from combining 50,000- and 777,000-marker genotypes from four countries. Interbull Annual Meeting Proceedings. Cork, Ireland, May 28–June 1, 4 pp. Interpretive Summary: Genomic predictions were compared on U.S. scale after combining genotypes from bulls from four countries. Reliabilities of predictions for younger animals improved 2.6% when foreign Holstein bulls were added to reference population. Many densities of single nucleotide polymorphism (SNP) genomic chips are now available, including 777,000 (HD), 56,000 (50K), and lower density including 3,000 (3K), 6,000 (6K) and 8,000 (8K) SNP chips. Using HD vs. 50K genotypes resulted in only a 0.4% gain in average reliability of genomic predictions. A nonlinear model with heavy-tailed priors for marker effects gave only a 0.8% increase over the linear model. Accuracy for 3K and 6K genotypes improved about 2% if imputed first to 50K and then to HD instead of imputing all genotypes together. International exchange continues to improve the reliability of evaluations and reduce cost of obtaining additional genotypes.
Technical Abstract: Genomic predictions were compared on U.S. scale after combining 50,000 (50K) and 777,000 (HD) marker genotypes across countries. The genotyped Holsteins included 161,341 animals with five marker densities including 1,510 with HD. Imputation was more accurate with FImpute than with findhap across the five densities. Reliability of HD predictions minus reliability of 50K predictions averaged only 0.4%. For 50K predictions, reliability for the combined reference population was higher than for domestic only, with the gain averaging 2.6% for Holsteins and 3.2% for Brown Swiss when 50K genotypes were added for 3,593 foreign Holstein and 732 Brown Swiss bulls that had no U.S. daughters prior to August 2008. Gains should be larger with 7,974 foreign Holstein and 870 Brown Swiss bulls in the May 2012 reference population. Multi-trait methods were not more accurate than single-trait for Holsteins, but multi-trait reliability averaged 1.4% higher than single-trait for Brown Swiss, perhaps because of lower genetic correlations between Brown Swiss populations. International exchange improves accuracy and decreases cost of obtaining additional phenotypes and genotypes.