Size of required reference population updates to achieve constant genomic prediction accuracy across generations

Abstract Text: High accuracy of estimated breeding values is crucial for achieving high genetic progress. The accuracy of genomic breeding values drops if the reference population is used over generations without supplementing it with new animals. The goal of this study was to investigate how many animals per generation need to be added to the reference population to keep the accuracy at a constant level across generations. On average the accuracy dropped by 0.07 when moving from first to second generation. After adding 25% of the initial reference population size the accuracy returned to its initial value. The required number of animals that were added to the reference population varied substantially illustrating that probably there are better strategies than adding animals at random. A possible solution is to consider the relationships between the animals used to update reference population and the selection candidates.

Keywords: dairy cattle, reference population, accuracy, relationships