Genomic and genetic evaluation of feed efficiency and stillbirth in dairy cattle

Feed efficiency is an economically important trait in the dairy cattle industry, and feed costs accounts for more than 50% of total production costs. Prediction of genetic breeding value has been a focus of animal breeding since the beginning of the 20th century. Because of ongoing genetic selection for productivity and improvement in herd management, the efficiency of converting feed to milk in U.S. dairy cattle has doubled over the past 60 years due to dilution of maintenance. It is widely recognized that additional selection based on biological differences between individuals in feed efficiency is highly desirable. The emergence of high dimensional genomic data offers opportunities for selection and evaluation of feed efficiency directly through whole genome-enabled prediction. This thesis centers on genetic evaluation and prediction of traits related to feed efficiency in dairy cattle using whole genome molecular markers. We investigated various whole genome prediction approaches tailored to capturing total genetic variation, with the goal of enhancing predictive performance for feed efficiency and related traits. In particular, this thesis includes three studies. In the first study, a semi-supervised learning approach was introduced, and its prediction accuracy was assessed using residual feed intake (RFI) data. The second study compared an interaction model with within- and across-environment components using data from multiple environments to estimate genomic variances and assess the accuracy of genomic predictions for RFI and its component traits. The third study involved genetic evaluation of direct and maternal stillbirth rate, a trait that contributes to whole farm production efficiency, using data of Brown Swiss, Jersey, and Holstein bulls. Our results indicate that, while selection on feed efficiency in dairy cattle using whole genome molecular markers is promising, low accuracy of prediction remains an ongoing challenge due to the limited size of the reference population. Pooling data across countries or production systems is an option for increasing size of the reference population, but genotype by environment interactions and population stratification must be addressed. Ongoing collection of individual feed intake records is necessary to improve prediction accuracy, in terms of increasing the size of the reference population and ensuring that reference animals are closely related to the current selection candidates.