Genomic Selection in Perennial Ryegrass: Results from Empirical and Simulation Studies

Genomic selection aims to predict selection candidates with only genotypes from marker effects estimated in a reference population comprised of genotyped and phenotyped individuals.  The main factors affecting the accuracy of genomic selection are reference population size, trait heritability, the proportion of the genetic variance captured by the markers, and the effective number of chromosome segments (M_e).  Breeding populations with a large M_e, such as perennial ryegrass, are diverse and also need large reference populations to achieve moderate accuracy of genomic selection.  However, breeding programs that reduce genetic diversity substantially have been proposed.  Within these closely related populations moderate to high accuracy is possible even with modest reference population size, regardless of original population diversity.  In the extreme, this means the application of full and half-sib genomic selection designs, which makes reference populations very specific and may require their complete renewal in each breeding cycle.  We have applied genomic selection at the cultivar level (i.e. less related than full and half-sib) in both empirical and simulated genomic selection studies.  In our empirical field trial data and using genetic markers from candidate genes as well as a transcriptome-based genotyping-by-sequencing approach, the accuracy of genomic selection was moderate to high within cultivar for heading date but it was reduced for biomass yield.  Our simulations of commercial perennial ryegrass breeding programs show that genetic gain per year can be at least doubled for moderate heritability traits (e.g. biomass yield) and significantly increased for low heritability traits (e.g. persistency).