Abstract
Abstract
Key message
Historic breeding program end-use quality data was used to perform genomic selection in two broad frameworks, incorporating fixed effect loci and genotype by environment interactions.
Abstract
Exceptional end-use quality in hexaploid bread wheat is an important factor for selling wheat in the global marketplace. However, many end-use quality traits are time-consuming and expensive to test, meaning their measurement is often delayed until late-stage trials of a breeding program. Genomic selection (GS) is one tool that has been shown to allow for earlier generation improvement of many traits in wheat breeding. Further, advances in GS methods may improve model predictive abilities, such as incorporating known loci that control traits as fixed effects and utilizing single-stage reaction norm models to control for genotype by environment interactions. Therefore, we evaluated conventional methods as well as these two advances to perform GS for twelve important end-use quality traits collected by the Montana State University (MSU) Spring Wheat Breeding Program from 2010–2022 across 842 elite breeding lines. We found that the conventional GS models performed well in 5-fold cross validation with predictive abilities ranging from 0.4–0.7 and in leave-one-year-out validation with most trait-year combinations exceeding a predictive ability threshold of 0.3. We did not observe any difference in predictive ability for the inclusion of fixed effect loci, but we did observe strong within-year correlations for the single-step reaction norm models in both validation methods. We concluded that both conventional, two-step GS and single-step reaction norm GS offer viable methods for earlier generation improvement of major end-use quality traits in the MSU Spring Wheat Breeding Program.