Accuracy of Genome-Wide Prediction for Fusarium Head Blight Associated Traits in a Spring Wheat Breeding Program

The development of wheat varieties with resistance to Fusarium head blight is a major breeding goal for the University of Minnesota’s spring wheat breeding program.   Currently, the program selects for FHB resistance using a combination of targeted molecular markers associated with major QTL, and phenotypic evaluation in misted, inoculated disease nurseries over several years and environments.  Phenotypic selection for FHB resistance is a resource intensive activity, and adds time to the breeding cycle.  Genome-wide predictive modeling of line performance has the potential to reduce time and resources allocated to phenotypic selection.  Program and trait-specific models need to be evaluated to empirically assess their efficacy in predicting the performance of lines for FHB associated traits.  To assess accuracy of genome-wide prediction models for the UMN Wheat Breeding Program, a panel of 383 F₇-derived lines were phenotyped in a minimum of five environments between 2009 and 2013.  The panel was genotyped at high density using the 90K Illumina Infinium iSelect Assay, resulting in 16,697 SNP markers for model training and prediction.  Here we present genome-wide prediction model accuracies for FHB-related traits, expressed as the correlation between the phenotype predicted by the model, and the BLUP of the observed phenotype.