Nonparametric methods for incorporating genomic information into genetic evaluations: an application to mortality in...

Area of Science:

• Animal Genetics
• Quantitative Genetics
• Bioinformatics

Background:

• Genetic evaluation of livestock relies on accurate heritability estimates.
• Traditional methods like E-BLUP may not fully capture complex genetic architectures influencing traits like mortality.
• Genomic information offers potential to enhance genetic prediction accuracy.

Purpose of the Study:

• To compare the performance of genomic-assisted genetic evaluation methods against a standard E-BLUP approach.
• To assess the utility of single-nucleotide polymorphism (SNP) information in predicting broiler sire mortality.
• To identify the most effective genomic approach for improving genetic evaluations in poultry.

Main Methods:

• Bayesian framework incorporating single-nucleotide polymorphism (SNP) data.
• Comparison of E-BLUP with four SNP-based methods: F(infinity)-metric model, kernel regression, RKHS regression, and Bayesian regression.
• Cross-validation was used to assess predictive ability.

Main Results:

• RKHS regression and kernel regression showed better data fitting with lower residual sum of squares.
• RKHS regression demonstrated the highest predictive accuracy, increasing it by 25-150% over other methods.
• Heritability estimates for mortality were low (0.02) with E-BLUP, indicating room for improvement with genomic data.

Conclusions:

• Genomic-assisted evaluation, particularly RKHS regression, significantly enhances the accuracy of genetic evaluations for broiler mortality.
• Nonparametric methods like RKHS regression are effective in utilizing SNP information for complex traits.
• Incorporating molecular markers holds substantial promise for improving genetic selection in poultry populations.