Exploration of Genomic Selection Options for a Commercial Perennial Ryegrass Breeding Program

Traditional ryegrass breeding programs have long breeding cycles (10-14 years), with multiple crossing and phenotypic selection stages (individual and swards/plots). Previous simulation work has shown that replacing phenotypic selection with genomic selection (GS) could substantially increase genetic gain for plot traits (e.g. yield), through reduced cycle time and the selection of individual plants using genomic breeding values (GEBVs) predicted from a plot reference population. However, GS also resulted in higher inbreeding than phenotypic selection. In this study we expanded the number of simulated GS scenarios to 9, aiming to balance genetic improvement, inbreeding rate and cost efficiency. Selection accuracy and intensity are factors affecting genetic gain and inbreeding. We simulated three plot reference population sizes (N_plots 100, 200, 500) and individual plant numbers selected based on GEBVs (N_individuals 2000, 5000, 10000). Cost efficiency was approximated as the breeding program cost of achieving 1kg/ha/year extra utilised yield for dairy farmers.

All genomic scenarios delivered more genetic gain for yield than conventional breeding (up to 4 times). Scenarios with larger N_plots and N_individuals contributed significant extra genetic gain and increased cost. Plots were especially costly and the extra gain from recruiting more plots in reference was generally not worth the extra expense. Thus, 10,000N_individuals&100N_plots and 5,000N_individuals&100N_plots were the most cost effective scenarios with inbreeding levels of 0.092 and 0.067, respectively, which could potentially affect sustainable long-term genetic gain. Overall, scenarios that genotyped moderate to high numbers of selection candidates, while maintaining moderate plot numbers, seemed most optimal in this breeding program design.