Search

Genomic prediction (GP) is a novel tool used for prediction of EBVs by using molecular markers. Within the last decade, GP has been widely introduced into routine evaluations of cattle, pig and sheep populations, however, its application in poultry has been somewhat delayed, and studies published to date have been limited in terms of population size and marker densities. This study shows a thorough evaluation of the benefits that GP could bring into routine evaluations of broiler chickens, with particular attention given to the accuracy and bias of Genomic BLUP (GBLUP) predictions. The data used for these evaluations exceeds the numbers of both individuals and marker genotypes of previously published reports, with the studied population consisting of up to 23,500 individuals, genotyped for up to 600K SNPs. The evaluation of GBLUP is preceded by evaluation of the variance components using traditional restricted maximum likelihood (REML) approach sourcing information from phenotypic records and pedigree, which provide an up to date reference for the estimates of variance components. Chapter 2 tested several models exploring potential sources of genetic variation and revealed the presence of significant maternal genetic and environmental effects affecting several commercial traits. In Chapter 3, a vast dataset containing 1.3M birds spread over 24 generations was used to evaluate changes in genetic variance of juvenile body weight and hen housed production over time. The results showed a slow but steady decline of the variance. Chapter 4 provided initial estimates of the accuracy and bias of genomic predictions for several sex-limited and fitness traits, obtained for a moderately sized population of over 5K birds, genotyped with 600K Affymetrix Axiom panel from which several chips of varying marker densities were extracted. The accuracy of those predictions showed a great potential for most traits, with GBLUP performance exceeding that of traditional BLUP. Chapter 5 investigated the effect of marker choice, with two chips used: one created from GWAS hits and second from evenly spaced markers, both with constant density of 27K SNPs. The two chips were used to calculate genomic relationship matrices using Linkage Analysis and Linkage Disequilibrium approaches. Markers selected through GWAS performed better in Linkage Analysis than in Linkage Disequilibrium approach. The optimum results however were found for relationship matrices which regressed the genomic relationships back to expected pedigree-based relationships, with the best regression coefficient dependent on the chip used. Chapter 6 formed a comprehensive evaluation of the utility of GBLUP in a large broiler population, exceeding 23,500 birds genotyped using 600K Affymetrix Axiom panel. By splitting the data into variable scenarios of training and testing populations, with several lower density chips extracted from the full range of genotypes available, the effect of population size and marker density was evaluated. While the latter proved to have little effect once 20K SNPs threshold was exceeded, the effect of the population size was found to be the major limiting factor for the accuracy of EBV predictions. The discrepancy between empirical results found and theoretical expectations of accuracy based on the similar genomic and population parameters showed an underestimation of the previously proposed requirements.

Soybean is a significant source of protein and oil, and also widely used as animal feed. Thus, developing lines that are superior in terms of yield, protein and oil content is important to feed the ever-growing population. As opposed to the high-cost phenotyping, genotyping is both cost and time efficient for breeders while evaluating new lines in different environments (location-year combinations) can be costly. Several Genomic prediction (GP) methods have been developed to use the marker and environment data effectively to predict the yield or other relevant phenotypic traits of crops. Our study compares a conventional GP method (GBLUP), a kernel method (Gaussian kernel [GK]), a machine learning method (deep learning [DL]) and a hybrid method that corresponds to the emulation of a machine learning model using a kernel method (and arccosine kernel [AK]) in terms of their prediction accuracies for predicting grain yield, oil and protein using data from the Soybean Nested Association Mapping experiment (1,379 genotypes tested in six environments, all genotypes in all environments). The relative performance of the four methods varied with the response variable and whether the model includes the genotype-by-environmental interaction effects or not. GBLUP consistently showed better performances, while GK and AK followed a similar pattern to GBLUP, DL performed slightly worse than the other three methods in most of the cases; however, this may also be attributed to sub-optimal hyperparameters. DL performed particularly worse than the other three methods in presence of the genotype-by-environmental interaction effects. In general, all four methods performed better when the interaction effects were included compared to the main effects only models. Advisor: Reka Howard