P.M. VanRaden*1, C.P. Van Tassell1,2, G.R. Wiggans1, T.S. Sonstegard2, R.D. Schnabel3, and F. Schenkel4
1Animal Improvement Programs Laboratory, Agricultural Research Service, USDA, Beltsville, MD, USA; 2Bovine Functional Genomics Laboratory, Agricultural Research Service, USDA, Beltsville, MD, USA; 3University of Missouri, Columbia, MO, USA; 4University of Guelph, ON, Canada
2008 J. Dairy Sci. (?)
© American Dairy Science Association,
2008. All rights reserved.
Individuals may download, store, or print single
copies solely for personal use.
Do not share personal accounts or passwords
for the purposes of disseminating this article.
Genetic progress will increase when breeders examine genotypes instead of only pedigrees and phenotypes. Genotypes for 39,835 markers and August 2003 genetic evaluations for 2609 Holstein bulls born before 1998 were used to predict January 2008 daughter deviations for 510 bulls born 2001-2. Genotypes were from the Illumina Bovine SNP50™ chip and semen contributed by U.S. and Canadian AI organizations to the Cooperative Dairy DNA Repository. Genomic predictions for 5 yield traits, 5 fitness traits, 16 conformation traits, and net merit were computed by a linear model with an assumed normal distribution for marker effects and also a nonlinear model with a heavier tailed prior to account for major genes. The official parent average from 2003 and a 2003 parent average computed from only the subset of genotyped ancestors were combined with the genomic predictions by selection index. The combined predictions were significantly (P < .0001) more accurate than official parent averages for all 27 traits. Squared correlations were 0.03 to 0.22 higher with linear genomic predictions included than those from parent average alone. Nonlinear genomic predictions had R-square similar to linear except for a few traits such as fat percentage, with maximum additional increase of 0.08. Squared correlations were converted to realized reliabilities by dividing by the average reliability of 2008 daughter deviations and by adding the difference between published and observed reliabilities of the 2003 parent averages. When averaged across all traits, combined genomic predictions had realized reliabilities 18% higher than reliabilities of parent average (48% vs. 30%), and gains in information were equivalent to 9 daughter records. Reliability increased more by doubling the number of bulls genotyped than the number of markers genotyped. Genomic selection can decrease generation interval and greatly increase accuracy by tracing the inheritance of minor genes.