Optimum Genotyping Strategy for Genomic Prediction By Using Association Study, a Case of Study in Wheat Advanced Breading Lines

In the last decade genomic selection (GS) has been implemented, especially for animals, as a new breeding technology. More recently, the opportunity offered by GS to speed up  genetic gains in complex traits without the need of identifying markers associations with the QTLs, have attracted attention of plant breeders. However, rather inexpensive and accurate plant phenotypes can often be obtained; hence a cost efficient genotyping method is pivotal for GS implementation in the breeding procedure.

In the present study 645 advanced wheat lines from standard breeding program of Nordic Seed Inc. were genotyped with the Illumina 15K wheat chip and phenotyped across different years and locations. After filtering 11045 SNPs were available and yield and lodging score were analyzed. K-fold cross-validation using all the markers in a standard gBLUP model showed predictive abilities (PA) for both traits around 0.55. Moreover PA where computed for progressively smaller, randomly selected SNPs subsets.  This exercise showed that around 1K markers were enough to reach a PA of 0.5.  An association study (AS) for both traits identified few significantly associated markers. When the selection of SNPs subsets was based on the significance level yielded by the AS an increase of genomic predictions was observed for small subsets and a PA of 0.5 was reached with 300-500 AS-selected SNPs.

Our results show limitations of AS when working with elite breeding lines because of small QTL effects. Nevertheless, AS can be used to increase the efficiency of GS by using the larger (mostly non-significant) marker effects.