Estimation of genomic breeding values for residual feed intake in a multibreed cattle population.

Khansefid, M., Pryce, J.E., Bolormaa, S., Miller, S.P., Wang, Z., Li, C., and Goddard, M.E. (2014). "Estimation of genomic breeding values for residual feed intake in a multibreed cattle population.", Journal of Animal Science, 92(8), pp. 3270-3283. doi : 10.2527/jas.2014-7375  Access to full text


Residual feed intake (RFI) is a measure of the efficiency of animals in feed utilization. The accuracies of genomic EBVs (GEBVs) for RFI could be improved by increasing the size of the reference population. Combining RFI records of different breeds is a way to do that. The aims of this study were to: 1) develop a method for calculating GEBVs in a multi-breed population and 2) improve the accuracies of GEBVs by using SNPs associated with RFI. An alternative method for calculating accuracies of GEBVs using genomic BLUP (GBLUP) equations is also described and compared to cross-validation tests. The dataset included RFI records and 606,096 SNP genotypes for 5,614 Bos taurus animals including 842 Holstein heifers and 2,009 Australian and 2,763 Canadian beef cattle. A range of models were tested for combining genotype and phenotype information from different breeds and the best model was the one including an overall effect of each SNP, an effect of each SNP specific to a breed and a small residual polygenic effect defined by the pedigree. In this model the Holsteins and some Angus cattle were combined into one "breed-class" because they were the only cattle measured for RFI at an early age (6-9 months of age) and were fed a similar diet. The average empirical accuracy (0.31), estimated by calculating the correlation between GEBV and actual phenotypes divided by the square root of estimated heritability in 5 fold cross-validation tests, was near to that expected using the GBLUP equations (0.34). The average empirical and expected accuracies were 0.30 and 0.31, respectively, when the GEBVs were estimated for each breed separately. Therefore, the across breed reference population increased the accuracy of GEBVs slightly, although the gain was greater for breeds with smaller number of individuals in the reference population (0.08 in Murray Grey and 0.11 in Hereford for empirical accuracy). In a second approach, SNPs that were significantly (P < 0.001) associated with RFI in the beef cattle GWAS were used to create an auxiliary genomic relationship matrix for estimating GEBVs in Holstein heifers. The empirical (and expected) accuracy of GEBVs within Holsteins increased from 0.33 (0.35) to 0.39 (0.36) and improved even more to 0.43 (0.50) when using a multi-breed reference population. Thus, a multi-breed reference population is a useful resource to find SNPs with a greater than average association with RFI in one breed and use them to estimate GEBVs in another breed.

Date modified: