Accuracy of genomic breeding values in multibreed beef cattle populations derived from deregressed breeding values and phenotypes1,2

Abstract

Genomic selection involves the assessment of genetic merit through prediction equations that allocate genetic variation with dense marker genotypes. It has the potential to provide accurate breeding values for selection candidates at an early age and facilitate selection for expensive or difficult to measure traits. Accurate across-breed prediction would allow genomic selection to be applied on a larger scale in the beef industry, but the limited availability of large populations for the development of prediction equations has delayed researchers from providing genomic predictions that are accurate across multiple beef breeds. In this study, the accuracy of genomic predictions for 6 growth and carcass traits were derived and evaluated using 2 multibreed beef cattle populations: 3,358 crossbred cattle of the U.S. Meat Animal Research Center Germplasm Evaluation Program (USMARC_GPE) and 1,834 high accuracy bull sires of the 2,000 Bull Project (2000_BULL) representing influential breeds in the U.S. beef cattle industry. The 2000_BULL EPD were deregressed, scaled, and weighted to adjust for between- and within-breed heterogeneous variance before use in training and validation. Molecular breeding values (MBV) trained in each multibreed population and in Angus and Hereford purebred sires of 2000_BULL were derived using the GenSel BayesCπ function (Fernando and Garrick, 2009) and cross-validated. Less than 10% of large effect loci were shared between prediction equations trained on (USMARC_GPE) relative to 2000_BULL although locus effects were moderately to highly correlated for most traits and the traits themselves were highly correlated between populations. Prediction of MBV accuracy was low and variable between populations. For growth traits, MBV accounted for up to 18% of genetic variation in a pooled, multibreed analysis and up to 28% in single breeds. For carcass traits, MBV explained up to 8% of genetic variation in a pooled, multibreed analysis and up to 42% in single breeds. Prediction equations trained in multibreed populations were more accurate for Angus and Hereford subpopulations because those were the breeds most highly represented in the training populations. Accuracies were less for prediction equations trained in a single breed due to the smaller number of records derived from a single breed in the training populations.

Genome Properties and Prospects of Genomic Prediction of Hybrid Performance in a Breeding Program of Maize

“…Weber et al (2012) used whole-genome estimates of marker effects of several cattle breeds to investigate across-breed marker-QTL linkage phase consistency. Such a study is still missing for maize and other important crops.…”

Genomic selection in multi-breed dairy cattle populations

“…They found that application of single-breed SNP effects estimates in other breeds performed quite poorly; estimates from a multi-breed population performed better than parent average but more poorly than single-breed estimates; and a multiple trait model in which breeds were treated as traits performed similarly to single-breed estimates. Weber et al (2012) also reported that accuracies were lower when using single-breed beef cattle reference populations, but that gains in accuracy were highest for the breeds best represented in the multi-breed training population.…”

Accuracy of genomic values predicted using deregressed predicted breeding values as response variables

“…Deregression of EBVs adjusts for ancestral information, it removes shrinkage present in EBV, and by taking parental contribution into account, DEBVs can be regarded as equivalent to the information provided by the records of each sire and its progeny (13) .…”

“…Accuracies (R 2 ) of GVs were obtained as the square of the correlation between GVs and the true genetic values (6,13,15) . Criteria for comparing the two alternatives of analysis were R 2 and GV prediction error variance (PEV).…”

“…The results of a study with two multi-breed beef cattle populations and Angus and Hereford purebred populations (13) used to obtain the GVs and corresponding R 2 for six growth and carcass traits showed that accuracies were lower for prediction equations trained in a single breed. These results were attributed to the smaller number of records derived from a single breed in the training populations.…”