Using Genomic Prediction to Characterize Environments and Optimize Prediction Accuracy in Applied Breeding Data

Simulation and empirical studies of genomic selection (GS) show accuracies sufficient to generate rapid annual genetic gains. Whole-genome genotyping provides the opportunity to go beyond the evaluation of lines to the evaluation of alleles and thus, provides new tools to analyze multienvironment trials. Considering allele replication rather than line replication provides a new way to cope with highly unbalanced phenotypic data sets. Using a two-row elite barley (Hordeum vulgare L.) population representative of the type of data generated by a commercial breeding program and tested for grain yield across Europe from 2007 to 2010, we characterized allele effect estimates at each test location and used them to successfully identify outlier environments. We also used the prediction accuracy between environments to characterize the environments. The prediction accuracy gave the same pattern as the genetic correlation between environments based on a factor analytic model, suggesting that it could be used to cluster environments. A new method was developed to optimize the composition of the training population for predicting performance in the target population of environments (TPE). This method does not search for mega-environments, but instead it identifies and removes less predictive environments from the set of environments used to train the model. Using this approach with the barley data set, cross-validated accuracy increased from 0.54 to 0.61 while controlling overfitting and focusing the prediction on the TPE. This study demonstrates the possibilities offered by GS to analyze multienvironment trials, identify outliers, group environments, and select historical data relevant for current breeding efforts.