The Accuracy Obtained from Reference Populations for Genomic Selection

For the design of breeding programs it is important to understand how trait measurement translates into selection accuracy. The introduction of genomic selection has created new challenges, in particular in relation to designing reference populations and valuing information sources for their contribution to genetic gain. The accuracy of genomic prediction depends on trait heritability, the number of phenotypes used (on genotyped animals) and the ‘effective number of chromosome segments’ that need to be estimated. The latter parameter is challenging to estimate but can in principle be derived from the variation in relationships between the reference set and the target animal. This paper attempts to validate that theory based on real data, with the aim to develop further insight into the value of a certain reference set for the genomic prediction of a certain target animal.