Quantifying genomic connectedness and prediction accuracy from additive and non-additive gene actions

Area of Science:

• Quantitative Genetics
• Genomic Prediction
• Animal Breeding

Background:

• Genetic connectedness is crucial for accurate breeding value comparisons across populations.
• Whole-genome prediction requires robust connectedness measures between reference and validation sets.
• The impact of non-additive gene action on genomic connectedness and prediction accuracy remains underexplored.

Purpose of the Study:

• To assess how non-additive kernel relationship matrices influence genomic connectedness measures.
• To investigate the relationship between connectedness and prediction accuracy under non-additive genetic variation.
• To extend the concept of genetic connectedness to include complex genetic architectures.

Main Methods:

• Utilized best linear unbiased prediction (BLUP) and coefficients of determination.
• Analyzed simulated data with additive, dominance, and epistatic effects.
• Applied non-additive kernel and non-parametric relationship matrices to real swine data.

Main Results:

• Non-additive kernel matrices significantly increased connectedness measures by up to 25%.
• Joint use of additive and non-additive matrices improved prediction accuracy compared to additive-only approaches.
• Enhanced connectedness correlated with improved predictive performance in the presence of non-additive gene action.

Conclusions:

• Genomic connectedness metrics can effectively incorporate non-additive genetic variation for complex traits.
• Kernel methods capturing non-additive effects enhance both connectedness and whole-genome prediction accuracy.
• This study broadens the application of genomic connectedness in genetic evaluations.