Files
Abstract
Proper evaluation of animals in breeding programs is essential to maximize response to selection. While breeding programs continually refine their selection strategies, they face evolving challenges stemming from factors such as the increasing number of genotyped individuals and reduced additive genetic variance due to selection. These challenges are even more pronounced for breeding programs for poultry species due to their rapid generation intervals, which speed up selection responses compared to other livestock species. This dissertation aims to address some of the current challenges affecting the breeding programs of broiler breeders. Accurate genetic parameters are vital for predicting breeding values to maximize the selection response. Changes in genetic parameters and correlations of SNP marker effects were estimated over time to investigate the effect of genomic selection on the population. Under genomic selection, the decline in additive genetic variation is faster, and thus, genetic parameters must be updated frequently with all the information used for selection decisions. In genomic evaluations, it is imperative to make selection decisions using the correct trait definition to ensure continued genetic gains. In the poultry industry, mortality is an economically important trait with financial and societal pressure for improvement. Alternative trait definitions were investigated for chicken mortality to explore whether any model accuracy improvement existed or whether a maternal genetic effect could be applied to this trait. Splitting mortality into time periods with the inclusion of the maternal genetic effect for early mortality may enhance the genetic evaluation in broiler breeder populations compared to the trait definition currently used for evaluations. Like mortality, many traits evaluated in breeding programs are binary or categorical by nature. Evaluating categorical traits can be costly regarding both time and computing performance, especially with large genomic data. A new method using an EM algorithm was investigated for efficiency without sacrificing breeding value accuracy within categorical trait evaluations. This approach used phenotypes and residuals from a linear model to impute liabilities for the categorical traits.