Predicting breeding values and accuracies from group in comparison to individual observations

Area of Science:

• Animal breeding and genetics
• Quantitative genetics
• Livestock production

Background:

• Individual animal observations are standard in livestock genetic evaluations.
• Pooled data, where only group performance is measured (e.g., pen feed intake), presents unique challenges.
• Traditional genetic evaluation methods may require adaptation for pooled data scenarios.

Purpose of the Study:

• To investigate the accuracy of genetic evaluations using pooled livestock data.
• To compare an exact mixed model method with an approximate method for handling pooled observations.
• To assess the impact of pool size and pooling strategy on evaluation accuracy.

Main Methods:

• Simulated a 3-generation livestock dataset with feed intake traits.
• Applied an exact mixed model approach, modifying variance-covariance matrices for pooled data.
• Implemented an approximate method treating pooled averages as individual records.
• Quantified empirical accuracy through product-moment correlation between true and estimated breeding values.

Main Results:

• The exact method yielded theoretical accuracies closely matching empirical results.
• Pooling data reduced empirical accuracies, with greater reductions for larger pool sizes and random allocation.
• The approximate method overstated accuracy and is not recommended.
• Selection effectiveness with pooled data was comparable to individual data for small pool sizes.

Conclusions:

• An exact mixed model approach provides accurate genetic evaluations with pooled livestock data.
• The exact method is computationally comparable to conventional procedures.
• Careful consideration of pooling strategy and size is crucial for maintaining evaluation accuracy.