Find information on spaces, staff, and services.
Prediction of genetic values has been a focus of quantitative genetics since the beginning of the 20th century, with renewed interest following the advent of the era of whole genome-enabled predict...
Prediction of genetic values has been a focus of quantitative genetics since the beginning of the 20th century, with renewed interest following the advent of the era of whole genome-enabled prediction. Opportunities offered by the emergence of high-dimensional genomic data fuelled by post-Sanger sequencing technologies and especially molecular markers, have driven researchers to extend Fisher's infinitesimal model to confront newly arising challenges. In particular, kernel methods are gaining attention as the regression method of choice for genome-enabled prediction. Complex traits are presumably influenced by genomic regions working in concert with others, thus generating interactions. Motivated by this fact, a growing number of statistical approaches based on kernels attempt to capture non-additive effects, either parametrically or non-parametrically. This thesis centers on whole-genome regression using kernel methods applied to a wide range of quantitative traits of agricultural importance in animals and plants. We investigated various kernel-based approaches tailored to capturing total genetic variation, to arrive at an enhanced predictive performance of complex traits in the light of available genome annotation information. In particular, this thesis reports on three studies conducted using kernel methods. In the first study using dairy cattle and wheat data, we constructed a diffusion kernel and compared its predictive performance against that of a Gaussian kernel. The second study evaluated some parametric and nonparametric kernels for predicting pre-corrected phenotypes and progeny tests in dairy cow health traits. The third study partitioned SNPs based on annotation and examined sources of predictive performance of complex traits in broiler chickens. Overall, while we obtained some encouraging results with non-parametric kernels, recovering non-additive genetic variation in a validation dataset still remains an ongoing challenge in quantitative genetics.