Genomic Selection for Wheat (Triticum aestivum L.) Trait and Trait Stability

Wheat yield, test weight, heading date, plant height, quality traits and trait stability are important for wheat breeding. Our objectives are to assess the relative efficiency of genomic selection (GS) for these traits and trait stability using four different models. Genotyping was conducted with a 90K SNP chip panel. SNP tagging was used to obtain a subset of 3,919 relatively independent markers for downstream analysis. Phenotyping was conducted on a population consisting of 273 lines, from seven different soft red winter wheat breeding programs in the U.S. Eberhart and Russell’s regression and Additive Main effects and Multiplicative Interaction (AMMI) models were used to assess trait stability. GS accuracy (r) was assessed for ridge regression Best Linear Unbiased Prediction (rr-BLUP), Bayesian Ridge Regression (BRR), Reproducing Kernel Hilbert Space (RKHS), and Elastic Net (EN). Across all models, accuracies were highest for test weight (0.61 to 0.67) and lactic acid content (0.63 to 0.65). GS produced a wide range of accuracies for trait stability (0.1 to 0.65) that varied by trait and stability method. Accuracy was 0.35 for yield and 0.44 for yield stability using AMMI, indicating the viability of GS in selecting lines with both high and stable yield. EN showed with higher accuracy than rr-BLUP for 11 out of 26 traits and trait stability index, and thus is recommended to use. BRR was not better than rr-BLUP except for yield. These findings lay the foundation for the wheat breeders in the Northeastern U.S. to implement GS into their breeding programs.