Comparison of genetically and genomically estimated variance

Last changed: 11 March 2013

Loberg, A.2, Crooks, L. 3 , Fikse,W. F.2, Dürr, J.W.1,2 and Jorjani, H.1,2, 1 Interbull Centre, P.O. Box 7023, SE-750007 Uppsala, Sweden, 2SLU, Dept of Animal Breeding and Genetics, Box 7023, SE-75007 Uppsala Sweden, 3 Wellcome Trust Sanger Institute, Cambridge, UK

To reach desired goals in animal breeding, the right breeding animals have to be selected. To find superior animals, estimation of genetic variance for the trait in question is fundamental, on national as well as international level. In this study estimation methods have been compared, by comparing the amount of genetic variance estimated. Data for 7038 Brown Swiss bulls have been included in this study. The phenotypic records included nationally and international estimated predicted genetic merits (NPGM, IPGM) for a production, a health, a conformation and a fertility trait. The genomic records included 44826 SNPs (after different quality checks). The data represent seven geographical regions (countries).

Five ways of estimating genetic variance were included in this study a) Estimates from each geographic region using phenotypic records on cows (using various REML implementations); b) Estimates from simple variance of the available NPGM values;  c) Estimates from an international  EM-REML implementation, d) Estimates using allele frequencies and allele effects in an international genomic evaluation model with  a non-linear,  heavy-tailed prior for marker effects analogous to Bayes A); and e)Genomic estimation of genetic variance with an AI-REML implementation. Results from method (a) were used as a reference base.  The amount of genetic variance estimated with different methods showed much variation within and across traits, also did the depth of pedigree affect results and will be investigated further.