|Weber, Kristina -|
|Van Eenennaam, Alison -|
Submitted to: Plant and Animal Genome Conference Proceedings
Publication Type: Abstract Only
Publication Acceptance Date: November 5, 2010
Publication Date: January 15, 2011
Citation: Weber, K., Bennett, G.L., Keele, J.W., Snelling, W.M., Thallman, R.M., Van Eenennaam, A., Kuehn, L.A. 2011. Genomic selection in beef cattle: Training and validation in multibreed populations [Abstract]. Plant and Animal Genome XIX Conference. Poster No. P514. Technical Abstract: A challenge for applying genomic selection to beef cattle is accurate across-breed prediction. One approach is to train and validate prediction equations in multibreed populations, but the scarcity of large populations with known pedigrees, phenotypes, and dense genotypes has hindered the development of across-breed genomic estimated breeding values (GEBV). Two populations that meet these requirements are the crossbred Germplasm Evaluation Program at the US Meat Animal Research Center (USMARC) and the purebred industry bulls of the 2000 bull project (2000_Bull), both genotyped using the BovineSNP50 BeadChip (50K) assay. However, it has not been established whether training on either dataset is sufficient to accurately predict genetic merit. To investigate this, a Bayesian method was used to predict GEBV for growth and carcass traits by training and cross-validation using observed phenotypes from 3358 USMARC cattle representing 8 breeds, and deregressed breeding values from 1937 of the 2000_Bull representing 12 breeds. Accuracies were calculated as the genetic correlation between GEBV and phenotypes within each population. For birth weight and weaning weight, the accuracies of 2000_Bull-trained GEBV were substantially higher than USMARC-trained GEBV, but for yearling weight and carcass traits, 2000_Bull-trained GEBV were the same or less accurate than USMARC-trained GEBV. Overall, GEBV accuracy ranged from 0.12-0.46 for 2000_Bulls-trained and from 0.22-0.37 for USMARC-trained. Different patterns of linkage disequilibrium were observed within each population. These results suggest that larger training populations and denser genotyping panels will be required to achieve higher accuracy across-breed predictions.