Genomic prediction in a backcross population using relationship matrices.

Area of Science:

• Quantitative genetics
• Animal breeding
• Genomic prediction

Background:

• Backcrossing is crucial for trait transfer in breeding.
• Traditional genomic selection models like G-BLUP simplify genetic architecture, potentially reducing prediction accuracy.
• Accurate genomic estimated breeding values (GEBVs) are vital for efficient genetic improvement.

Purpose of the Study:

• To develop and evaluate novel genomic prediction models that account for complex genetic architectures.
• To improve the accuracy of GEBVs compared to conventional G-BLUP.
• To provide more effective tools for accelerating genetic gain in breeding programs.

Main Methods:

• Developed three new models: covariance-adjusted genomic BLUP (CAG-BLUP) for correlated markers, and genomic-architecture-specific BLUP variants (GASI-BLUP for independent markers, GASC-BLUP for correlated markers) assuming unequal variances.
• Evaluated model performance using simulated and empirical mouse datasets.
• Compared the novel models against the traditional G-BLUP.

Main Results:

• GASI-BLUP significantly outperformed G-BLUP for independent quantitative trait loci (QTLs), increasing GEBV accuracy by up to 12% and reducing genetic variance underestimation.
• CAG-BLUP improved GEBV accuracy by up to 2% in scenarios with dependent QTLs, especially at lower heritabilities.
• The developed models captured genetic-architecture nuances more effectively than G-BLUP.

Conclusions:

• Selecting genomic prediction models tailored to specific trait genetic architectures is essential for enhancing prediction accuracy.
• The novel models (CAG-BLUP, GASI-BLUP, GASC-BLUP) offer superior GEBV prediction and better capture genetic nuances.
• These advancements can lead to more effective breeding programs, accelerating genetic improvement and contributing to food security.