Genomic selection for traits of economic importance in sheep : a thesis presented in partial fulfilment of the requirements for the degree of Doctor of Philosophy in Animal Breeding and Genetics at Massey University, Manawatū, New Zealand
Open Access Location
The main objective of this thesis was to analyse the inclusion of genomic information of production traits into a multitrait sheep breeding programme evaluated for 20 years using deterministic and stochastic simulation models. The breeding objective was to reduce faecal egg score (FES), decrease yearling weight (YW) and increase 160 days lamb carcass weight (CW). The selection criteria included 160 days live weight (instead of CW) plus YW and FES. The first study developed a stochastic model selecting animals based on their individual breeding values estimated using best linear unbiased predictor (BLUP) procedure with a multitrait animal model. The model was validated using a deterministic multitrait selection index; obtaining similar prediction responses for breeding objective and selection criteria traits. The second study deterministically evaluated the inclusion of genomic information explaining different proportions of CW and YW genetic variances into a selection index. Under the same selection scheme a selection index having only genomic information obtained lower accuracies and genetic gains compared to a selection index considering phenotypic information. If shorter generation intervals are implemented, a selection index including phenotypic and genomic information explaining low proportions of the trait's genetic variance could achieve higher genetic and economic gains. The third study evaluated genetic responses of a stochastically modelled breeding flock selecting ewes based on BLUP estimated breeding values and selecting rams based on genomic breeding values (GBV) for CW. The fourth study evaluated accuracy of prediction of CW GBV using the same simulated model. Carcass weight GBVs were calculated in a validation population using single nucleotide polymorphism (SNP) effects from a training population. The further apart the genetic relationship between these two populations, lower the GBV accuracy. Resultant accuracies depended on the proportion of total genetic variance explained by genomic information and also the variance accounted by each SNP, therefore an appropriate GBV estimating method has to be chosen to achieve accuracies as high as possible. Stochastic models proved to be more versatile for managing data, also showing variation of the genetic responses. In contrast, deterministic models were faster regarding computer processing times. The study proved that a breeding programme combining GBV and BLUP estimated breeding values can increase genetic responses by selecting animals at early stages of life.
Sheep breeding, Sheep genetics, Sheep breeding models, Livestock simulation models