Open access publications in the SLU publication database

Abstract

The use of genomic information has become increasingly important in a breeding program. In a pig breeding program, where the final goal is an increased crossbred (CB) performance, the use of genomic information needs to be thoroughly evaluated as it may require a different strategy of what is applied in purebred (PB) breeding programs. In this thesis, I explore the use of genomic information for the genetic improvement of PB and CB pigs. I first focus on the identification of genomic regions affecting traits that are important to breeders. I identified two quantitative trait loci (QTL) regions for gestation length, one for Dutch Landrace on Sus scrofa chromosome (SSC) 2 and the other one for Large White on SSC5. I also fine-mapped and narrowed down the region of a previously detected QTL for androstenone level SSC6 from 3.75 Mbp to 1.94 Mbp. A tag-SNP of this fine-mapped region was further investigated and no unfavorable pleiotropic effects were found; indicating that using the studied marker for selection would not unfavorably affect the other studied traits. After that, the focus was changed to the application of genomic selection in pigs. Within-population predictions showed high accuracies, whereas across-population prediction had accuracies close to zero. Using combinations among Dutch
Landrace and Large White populations plus their cross showed that multi-population prediction was not better than within-population. The exception was when the CB pigs were predicted with records from both parental populations added to the CB training data. When using PB pigs to train CB ones, the predictive ability found indicates that selection in the PB pigs results in response in the CB ones. When assessing the source of information used to estimate the breeding values used as response variable, I showed that a more accurate prediction of CB genetic merit was found when training on PB data with breeding values estimated using CB performance than training on PB data with breeding values estimated using PB performance. I also studied the accuracy of using CB pigs in the training population to select PB for CB performance. Predictive ability when using CB phenotypes for training was observed, however, the accuracy was lower than using PB phenotypes in the training population. Lastly, I evaluate the inclusion of dominance in the model when using a CB training population. Results showed that accounting for dominance effects can be slightly beneficial for genomic prediction compared with a model that accounts only for additive effects.