Enhancing wheat genomic prediction by a hybrid kernel approach.

Area of Science:

• Agricultural Science
• Genetics
• Bioinformatics

Background:

• Genomic selection models are crucial for predicting breeding values in crops.
• Integrating pedigree and genomic data can improve prediction accuracy.
• Existing models often fail to capture complex genetic architectures.

Purpose of the Study:

• To enhance genomic prediction accuracy by combining pedigree (A) and genomic (G) similarity matrices.
• To explore the utility of non-linear methods, specifically kernel matrices, for genetic analysis.
• To develop and validate novel hybrid kernels (C and P) from the interaction of A and G matrices.

Main Methods:

• Utilized multiple wheat datasets for single- and multi-environment analyses.
• Developed five single-environment and five multi-environment models.
• Incorporated genotype-by-environment (G × E) interactions in multi-environment models.
• Introduced two novel symmetric kernels (C and P) derived from the interaction of genomic and pedigree matrices.

Main Results:

• The proposed models S5 and M5, incorporating kernels C and P, significantly improved prediction accuracy.
• Hybrid kernels captured additional, independent genetic variation beyond conventional matrices.
• The novel prediction models outperformed standard conventional models in most scenarios.
• Genomic models employing non-linear kernels demonstrated superior predictive ability compared to linear models.

Conclusions:

• The integration of pedigree and genomic data through novel hybrid kernels enhances genomic prediction.
• Non-linear kernel methods offer a powerful approach for capturing complex genetic variation in breeding programs.
• The developed models provide a more accurate and robust tool for wheat genetic improvement.