Some abstracts do not have video files because ASAS was denied recording rights.
783
Objective-oriented genomic relationship matrices
The advent of affordable high-density SNP genotyping platforms has boosted the implementation of genomic selection program in many livestock species. However, large reference populations are required to accurately compute genomic predictions of breeding value (GEBV). Combining information from (seemingly independent) separate populations has been highlighted as a beneficial strategy and methods to adjust the realized genomic relationship matrix (GRM) to accommodate the heterogeneity in allele frequencies have been proposed. Simulation studies based on real sequence data have shown the importance of using only variants as close as possible or identical to the causative mutations. Recently, we showed that data from reference populations from two distinct breeds can be merged to generate GEBV, provided the SNP used to build the GRM are carefully selected based on their significance and direction of the effect associated to the phenotype. We show that this approach can optimise the genomic correlation for the phenotype of interest in the two populations. We further show how a ‘hybrid’ GRM permits the linking of genotypic data of pooled DNA samples of commercial cattle pooled according to phenotype with individual DNA samples from animals available for selection. Our examples are concerned with beef cattle raised extensively in tropical and subtropical regions of Australia. We anticipate that the use of traditional ‘one-size fits all’ relationship matrices, based on pedigree information only, is coming to an end, and predict the time has come for ‘objective-oriented’ GRM purpose-built for a specific breeding objective.
Keywords: genomic selection, genomic predictions, genomic relationship matrix