Fig. 4.

a Population structure in the breeding panels indicated by principal components 1 and 2. b Spot blotch prediction accuracies across the breeding panels and subsets evaluated between 2014 and 2020 in Agua Fria using the fixed effects model (with selected spot blotch associated markers as fixed effects), the genomic best linear unbiased prediction (GBLUP) model and the GBLUP + fixed effects model (GBLUP + fixed effects). The prediction accuracies for random folds indicate the fivefold cross validation accuracies using random folds in the breeding panel lines, where 1347 random lines were predicted from a training population of the remaining 5,389 lines and in the breeding panel subset lines, 843 lines were predicted from a training population of the remaining 3373 lines. In predictions across breeding panels, all the seven breeding panels with 904–1011 lines were predicted from large training sets of 5725–5832 lines comprising all the other panels. Similarly, in the subsets of breeding panels, 443–704 lines were predicted from training sets of 3512–3773 lines, comprising all other subsets