Optimization of Training Population Under Population Structure in Genomic Selection

Training population set (TRS) design for genomic selection has attracted much attention in both animal and plant breeding in recent years because it is critical to the accuracy of the prediction models.  In a scenario where we have a diverse panel of genotypes that have been genotyped but not phenotyped, the first question that arises is how to select the best genotypes to create a TRS to build our statistical model for making predictions in the test set (TS). The purpose of this study was to compare the performance of different optimization criteria in the presence of population structure and to evaluate how population structure interacts with these criteria in the choice of the TRS. Five different TRS sampling algorithms, stratified sampling, mean of the Coefficient of Determination (CDmean), mean of Predictor Error Variance (PEVmean), stratified CDmean (StratCDmean) and random sampling, were evaluated for prediction accuracy in the presence of different levels of population structure on wheat and rice datasets. The wheat dataset showed mild population structure and CDmean and stratified CDmean methods showed the highest accuracies for all the traits except for test weight and heading date. The rice dataset had strong population structure and the approach based on stratified sampling showed the highest accuracies for all traits. These results indicated that population structure must be evaluated before optimization of the TRS, as the selection algorithm choose for optimization depends on the degree of the population structure effects.