Skip to main content
ARS Home » Northeast Area » Beltsville, Maryland (BARC) » Beltsville Agricultural Research Center » Animal Genomics and Improvement Laboratory » Research » Publications at this Location » Publication #303159
Title: Increasing predictive ability using dominance in genomic selection

Author
item SUN, CHUANYU - National Association Of Animal Breeders
item Vanraden, Paul
item Cole, John
item O'CONNELL, JEFFREY - University Of Maryland

Submitted to: World Congress of Genetics Applied in Livestock Production
Publication Type: Proceedings
Publication Acceptance Date: 4/21/2014
Publication Date: 8/17/2014
Citation: Sun, C., Van Raden, P.M., Cole, J.B., O'Connell, J.R. 2014. Increasing predictive ability using dominance in genomic selection. World Congress of Genetics Applied in Livestock Production. Vancouver, Canada, Aug. 17–22. 3 pp.

Interpretive Summary:

Technical Abstract: Dominance can be an important source of nonadditive genetic variance for many traits of dairy cattle. The role of dominance was investigated for Holstein and Jersey milk, protein, and fat yields; somatic cell score (SCS), productive life (PL), and daughter pregnancy rate (DPR). Additive and dominance variance components were estimated first, and then those estimates were used to estimate additive and dominance effects of single-nucleotide polymorphisms. The predictive ability between 3 models with both additive and dominance effects (MAD1, MAD2, and MAD3) and a model with additive effect only (MA) were assessed using 10-fold cross-validation. The MAD1 model estimated dominance values, and MAD2 estimated dominance deviations with a different dominance relationship matrix; MAD3 enlarged the data set by including cows with genotype probabilities that were derived from genotyped ancestors. Dominance from MAD1 accounted for 5 and 7% of total variance for Holstein and Jersey yield traits, respectively. Heritability estimates were lower for dominance and higher for additive genetic effects with MAD2 than with MAD1. For SCS, PL, and DPR, Holstein and Jersey dominance variances were very small. Mean correlations between estimated genetic effects and phenotypes for the 10-fold cross-validation showed that MAD1 and MAD2 increased prediction accuracy relative to the MA model for yield traits, but no difference was found for SCS, PL, and DPR. Compared with MAD1 and MAD2, MAD3 did not further improve prediction.