Response to Early Generation Genomic Selection for Yield in Wheat

Area of Science:

• Plant Breeding and Genetics
• Agricultural Science
• Quantitative Genetics

Background:

• Accelerating genetic gain in grain yield is crucial for wheat improvement.
• Early generation genomic selection (GS) offers a promising approach to enhance breeding efficiency.
• Accurate prediction of breeding values in early generations is essential for successful GS implementation.

Purpose of the Study:

• To investigate the effectiveness of GS in early generations for increasing genetic gain in wheat grain yield.
• To evaluate different statistical models for calculating genomic estimated breeding values (GEBVs) in early generations.
• To compare the predictive ability of various GEBV calculation methods for grain yield.

Main Methods:

• A training set of 1,334 elite wheat lines was used to generate GEBVs for grain yield.
• Three prediction methods were applied: Genomic Best Linear Unbiased Predictor (GBLUP), rrGBLUP with imputation, and Reproducing Kernel Hilbert Space (RKHS/Gaussian Kernel).
• Early generation (F2) GEBVs were used to select individuals for subsequent generations and to compare predictive abilities using F2:4 lines.

Main Results:

• GEBVs from different prediction methods in F2 individuals were not correlated.
• Experiment 1 showed no significant yield difference between genomically selected and conventionally selected lines.
• Experiment 2 revealed significant positive correlations between observed F2:4 line yields and F2 GEBVs from Gaussian Kernel (0.248) and GBLUP (0.195), but not rrGBLUP_imp.

Conclusions:

• Genomic selection holds potential for application in early generations of wheat breeding.
• The choice of statistical model for GEBV calculation is critical and may differ between early generations and inbred lines.
• Gaussian Kernel and GBLUP demonstrated predictive ability for grain yield in early generations, highlighting the importance of model selection.