Improving hybrid breeding efficiency in winter oilseed rape under sparse multi-environment testing

Area of Science:

• Plant breeding and genetics
• Quantitative genetics
• Genomic selection

Background:

• Genomic prediction is vital in plant breeding, but its effectiveness in sparse testing scenarios, particularly in hybrid breeding programs for winter oilseed rape (Brassica napus L.), requires further investigation.
• The robustness of genomic prediction models under reduced data availability is a critical concern for optimizing breeding strategies.

Purpose of the Study:

• To evaluate the impact of various data omission strategies, prediction schemes, and statistical models on the accuracy and consistency of genomic predictions.
• To identify optimal approaches for genomic prediction in winter oilseed rape breeding programs facing data limitations.

Main Methods:

• Assessed seven data omission scenarios across three traits with diverse genetic architectures.
• Utilized correlation between observed and predicted phenotypes and Cohen's κ to measure prediction accuracy and selection consistency.
• Compared multi-environment predictions with per-environment predictions and evaluated models incorporating pedigree-genomic relationship matrices.

Main Results:

• Genomic prediction accuracy and selection consistency decreased with increased data omission, though the two metrics showed moderate correlation.
• Balanced omission across environments and relationship-informed omission strategies enhanced prediction performance and stability.
• Models combining pedigree and genomic data improved predictions and allowed inclusion of non-genotyped parents.
• Multi-environment predictions using General Combining Ability (GCA) + Specific Combining Ability (SCA) models generally outperformed other schemes, with trait-specific variations.

Conclusions:

• Informed data omission strategies and judicious model selection are crucial for maximizing genomic prediction performance under data reduction.
• Relationship-based balanced omission offers a promising strategy for efficient and accurate genomic selection in breeding programs.
• Breeders can significantly influence crop performance by strategically designing training sets and selecting appropriate prediction models.