Combining datasets to routinely predict fusarium head blight resistance in a wheat breeding program

Abstract

AbstractFusarium head blight (FHB; Fusarium graminearum Schwabe) is a devastating fungal disease of wheat (Triticum aestivum L.) that can significantly reduce yield and grain quality. Datasets from different stages of field evaluation can be combined into a training population to predict FHB resistance. Our objective was to determine if FHB resistance among F5 lines can be predicted accurately with historical lines, parental lines, and a subset of F5 lines. Lines at the F5 and preliminary yield trial (PYT) stages in the University of Minnesota wheat breeding program were evaluated in two locations from 2016 to 2020 and were genotyped with 3679 single nucleotide polymorphism markers. Historical datasets with 368 to 3015 lines had predictive abilities of −0.01 to 0.20, whereas F5 subsets had predictive abilities of 0.04–0.32. Adding subsets of F5 lines to the historical datasets led to incremental improvements in predictive abilities in most cases, especially when the subset was selected via the pedigree or k‐means approach. The most effective training populations were those that contained a subset of 200 F5 lines chosen via the k‐means method, the F5 parents, and the PYT lines tested in the same year, with predictive abilities that were usually higher than that of the F5 subset. We have started to use such combinations of datasets to routinely predict FHB resistance of F5 lines in our breeding program.