Evaluating the Potential of Genomic Selection for Grain Yield and Yield Component Traits in South Dakota Winter Wheat

Study overview

Grain yield in wheat is a complex quantitative trait influenced by key yield components, including thousand kernel weight (TKW), kernel length (KL), kernel width (KW), and kernel area (KA). This thesis evaluated two genomic selection models, ridge regression best linear unbiased prediction (rrBLUP) and reproducing kernel Hilbert spaces (RKHS), for predicting grain yield and yield-component traits in the South Dakota State University winter wheat breeding program. KL, KW, and KA were measured on 1,391 unique breeding lines from advanced and preliminary yield trials at two South Dakota locations over the 2023 and 2024 growing seasons using an automatic Vibe QM3 grain analyzer. Pearson correlations showed significant positive associations between grain yield and all four kernel traits (r = 0.27-0.46), and five-fold cross-validation showed moderate-to-high predictive ability for kernel morphology traits, with kernel width reaching the highest predictive ability (0.54). Incorporating a representative core subset of preliminary yield trial lines into the training population improved predictive ability across most traits, suggesting kernel size traits are predictable enough to support early-generation selection in the breeding program.