Genomic selection across multiple breeding cycles in applied bread wheat breeding.

Area of Science:

• Agricultural Science
• Plant Breeding
• Genetics

Background:

• Genomic selection (GS) is crucial for crop improvement, but its application across multiple breeding cycles in bread wheat is understudied.
• Existing cross-validation studies often use uniform genetic material across environments, not reflecting real-world breeding progression.

Purpose of the Study:

• To evaluate GS accuracy across five bread wheat breeding cycles.
• To assess cross-validation bias within cycles and explore methods to enhance prediction accuracy.
• To investigate the impact of outliers on GS performance.

Main Methods:

• Estimated prediction accuracy for grain yield, protein content, and protein yield in 659 inbred lines over five breeding cycles.
• Assessed bias in within-cycle cross-validation and employed cycle-based cross-validation to mitigate bias.
• Investigated the influence of outlier cycles and environments on prediction accuracy.

Main Results:

• Heritability was highest for protein content, followed by grain yield and protein yield.
• Within-cycle cross-validation showed substantial bias for protein yield (17-712%) and protein content (8-86%).
• Cycle-based cross-validation yielded maximum prediction accuracies of 0.51 (protein content), 0.38 (grain yield), and 0.16 (protein yield).
• Removing outlier cycles improved grain yield prediction accuracy to 0.41, while outlier environments had minimal impact.

Conclusions:

• Genomic selection shows promise in bread wheat breeding but requires careful bias assessment across breeding cycles.
• Cycle-based cross-validation and outlier management are essential for reliable GS accuracy estimation.
• Predicting protein yield by combining genomic estimated breeding values for grain yield and protein content improved accuracy for this derived trait.