Autoregressive single-step test-day models for genomic prediction and GWAS in Portuguese Holstein cattle

• Ano de defesa: 2019
• Autor(a) principal: Silva, Alessandra Alves
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Universidade Federal de Viçosa
• Palavras-chave em Português: Bovinos de leite - Melhoramento genético; Autocorrelação; Genômica; Lactação; Genética e Melhoramento dos Animais Domésticos
• Link de acesso: https://locus.ufv.br//handle/123456789/27125

The milk traits are termed longitudinal traits because they are measured over time. Therefore, these traits need to be evaluated using appropriate statistical models to take into account the covariance structure that exists among the repeated records. Autoregressive test-day (AR) model for multiple lactations has been routinely used for the national genetic evaluation of dairy cattle in Portugal. Under this model, the animals’ permanent environment are assumed to follow a first order autoregressive process as a long-term (auto-correlations between parities) and a short-term (auto- correlations between test-day within lactations) effects, taking into account the non- genetic correlations due to the cows’ repeated performance. Currently, given the relevance of genomic prediction in dairy cattle, it is essential to include dense marker information in national genetic evaluations. Therefore, the general objective with this thesis was to evaluate the inclusion of genomic information in the AR test-day model for multiple lactations for better understand the genetic and genomic aspects of milk related traits in Portuguese Holstein cattle. Firstly, to perform the genomic evaluation under AR model we evaluated the imputation accuracy for Portuguese Holstein cattle using several commercially available SNP panels in different densities with a relatively small number of genotyped animals. Genotype imputation was feasible and may be advantageous to the National genomic evaluations. Thus, we analyzed the feasibility of applying the single-step GBLUP (ssGBLUP) to analyze milk yield using the AR (H-AR) model in Portuguese Holstein cattle. The use of H-AR increased the reliability and reduced the bias of GEBVs compared to traditional evaluation. Therefore, these results suggest that the ssGBLUP methodology applied to AR models is feasible and may be advantageous to the National genetic evaluations. With the anticipated increase in the number of genotyped animals (for example by including females), it is expected that the H-AR will provide even higher reliabilities especially for the young stock, thus contributing to the improvement in the genetic progress. In asecond step, we evaluated the feasibility of using a weighted ssGWAS methodology under a multiple lactations AR test-day model to find genomic regions associated with milk, fat, and protein yields, and score somatic cells (SCS). Genomic regions associated with the analyzed traits were also identified simultaneously throughout the lactations, to provide a better understanding of the genetic architecture for these traits. The findings described in this thesis will contribute to advance the knowledge about genomic prediction and GWAS for milk related traits. In addition, this thesis provides the first results about the inclusion of genomic information in AR models, which will be important for future national genetic evaluations. Keywords: Autocorrelation. Dairy cattle. Gene function. Genomic evaluation. Imputation. Multiple lactations.