Files
Abstract
Improving the methods for genomic prediction and genetic parameters will increase accuracy and decrease the generation interval, resulting in increased genetic gain. The objective of this dissertation was to introduce methods to improve the accuracy, efficiency, and understanding of genomic prediction and genetic parameter estimation in large populations. Simulated datasets and datasets from dairy cattle and pig populations were used to test and analyze the methods. The differences in bias, accuracy, and computing time using ssGBLUP were negligible when blending the genomic relationship matrix with different proportions of the identity or pedigree relationship matrix for genotyped animals. However, a new algorithm was introduced that reduced the bottleneck in computing time from approximately 2 hours to less than one second. The number of independent chromosomes in a population is assumed to be 4NeL, where Ne is the effective population size and L is the genome length in Morgans (M). A segment effect model was compared with the true accuracy to test if all genetic variation could be explained in 4NeL segments. Segment accuracies maximized at 4NeL but were not as high as the true accuracy, suggesting a more biologically reasonable definition for segments is needed. Using genomic information in heritability estimation in large populations is computationally expensive. Method R using genomics was compared with AI-REML with genomics and reduced computing time from 9.5 to 1.6 hours. However, the heritability estimates were not as precise and had large standard errors compared to the AI-REML estimates. Improvement in high-throughput phenotyping methods is also needed to incorporate this information into genetic evaluations and increase genetic gain. Behavior traits were recorded using digital phenotyping in a pig population. The data quality was analyzed, and genetic parameters of the behavior traits and relating behavior traits to production traits were estimated. The behavior traits analyzed had heritabilities ranging from 0.19 to 0.57 and had low to moderate genetic correlations with production traits. As the amount of phenotypic and genomic information is increasing rapidly, methods must be improved continuously to utilize the information and incorporate it into genetic evaluations.