Multi-population genomic prediction
In livestock breeding programs, genomic information is more and more used to select the genetically best animals to produce the next generation. For numerically small populations, the size of the population limits the accuracy of predicting genomic breeding values, which restricts the rate of genetic improvement. An attractive approach to increase the accuracy for those populations is to add information from other populations. This thesis aimed to investigate the accuracy of genomic prediction using information from multiple populations, by 1) investigating the effect of different parameters on the accuracy, and 2) by deriving deterministic equations to predict the accuracy as a function of those parameters. Results showed that absence of close family relationships, differences in the relation between markers and genes, differences in gene effects, and differences in the properties of genes across populations reduce the potential to use information from other populations. Moreover, two equations were derived to predict the accuracy of genomic breeding values. Based on those equations, it was shown that it is only beneficial to use information from other populations when the populations are closely related, the population itself is small, and a large number of animals from another population is added.