Accuracy and Training Population Design for Genomic Selection on Quantitative Traits in Canadian Durum Wheat Breeding Lines

Yield and quality traits (particularly gluten strength), are important selection criteria in durum wheat variety development. Due to the complexity of the inheritance of these traits, traditional marker assisted selection techniques are not effective.  Genomic selection (GS) could be a viable alternative for selection where predicted breeding value is the basis for selection. We used 196 advanced Canadian breeding lines to assess the possible utility of GS for yield and gluten strength using available genotypic data collected using the high density Infinium iSelect 90K SNP assay. Our objectives were to (i) assess the effects of changing training population size on prediction accuracy and (ii) assess accuracies when accounting for population structure and relatedness of lines within the training population. A total of 12,234 bi-allelic markers spanning all 14 durum wheat chromosomes were scored with a minor allele frequency >5%. Accuracy improved as the number of lines within the training population increased. Including older “founder” lines in the training population increased or maintained accuracy, indicating that older generations contained information useful for predicting breeding values. The presence of population structure affected accuracy: when training and validation subpopulations were closely related accuracy was greater than when they were distantly related. The low accuracies revealed for yield may be due to small population size but were still comparable with the results of published studies using a similar GS model. Overall, levels of prediction accuracy found in this study indicate that GS could be applied to facilitate selection in a durum wheat breeding program.