Genomic selection for high quality beef production

Abstract

This thesis focuses on the implementation of genomic selection within Wagyu, a breed of cattle that is highly desired due to its propensity to accumulate marbling. Initial focus of the thesis was to investigate using genomics to breed purebred Wagyu, producing crossbreds with improved marbling performance. However, the thesis had to undergo a change in direction due to unforeseen delays in obtaining crossbred genotype and phenotype data. The experimental chapters, therefore, focus on scenarios within the core nucleus breeding herd while the literature review considers the influence of crossbreeding heavily. Chapter two considered a comparison between pedigree and genomics with relationship matrices built from 10,549 and 4,940 individuals respectively. Animal models for multiple traits found genomics resulted in more accurate breeding values. This was evident through higher breeding value standard deviations and lower mean breeding value standard error. Additionally objective carcass measures were more heritable than subjective measurements (Meat Image Japan (MIJ) vs. AUS-MEAT grading) and highly correlated to their equivalent AUS-MEAT counterparts. This is consistent with findings from the meta-analysis in the literature review. Chapter three investigated how differing SNP densities describe genomic relationships across the Wagyu population herein, utilising masked subsets from a 30K base SNP density and HD SNP data. It was demonstrated that small SNP subsets of 2,500-5,000 were sufficient. Imputation was used to impute these subsets to a ~30K density, producing a genomic relationship matrix (GRM) with highly correlated elements to a GRM built using all 30K SNP data. Imputation to a high density SNP platform (770K) improved the description of relationships further by better describing highly related animals. Given imputation requires well-formed reference populations, Chapter four compared four published methods to select animals to form a reference population for imputation to whole genome sequence. Methods investigated used relationship matrices or haplotype libraries. The MCG method, which utilises a genomic relationship matrix to select animals highly related to the target population but distantly related to other selected candidates, accounted for the most genetic variance in the population relative to the other methods when 100 animals were selected. This method was then used to select 70 animals to be sent for whole genome sequencing. Chapter 5 planned to compare genetic parameters estimated from imputed whole genome sequence data to those from the commercial 30K SNP chip. Due to delays in obtaining sequencing data, a back-up set of 770K genotypes that accounted for a similar proportion of genetic variance as the original 70 animals selected by MCG (Chapter 4), was used to build an HD genomic relationship matrix to compare trait heritabilities and animal selection decisions. Animal models were used, as in Chapter 2, finding HD genotype arrays resulted in improved prediction accuracy through increased spread of breeding values and higher heritability estimates across traits. With Wagyu product worth an exceptional premium and with multiplier effects of genetic gain from the nucleus to daughter herds, marginal gains in accuracy are of high value. This supports that investment in higher density genotyping, including sequencing, and objective marbling assessment.