Predicting observed and latent responses to BLUP selection involving logistic variates

Area of Science:

• Animal breeding and genetics
• Quantitative genetics
• Statistical genetics

Background:

• Multivariate Best Linear Unbiased Predictors (BLUPs) are essential for animal breeding.
• Handling mixed continuous and discrete traits in BLUPs presents computational challenges.
• Logistic latent traits influence discrete trait expression in breeding populations.

Purpose of the Study:

• To develop and evaluate algorithms for predicting responses to BLUP selection with mixed trait data.
• To assess the accuracy of Taylor expansions compared to quadrature methods for complex breeding schemes.
• To provide efficient prediction methods for breeding programs involving discrete traits.

Main Methods:

• Development of algorithms using Taylor expansions for predicting discrete and latent responses.
• Re-parametrization of models to incorporate probabilities and moments of truncated multinormal distributions.
• Comparison of Taylor expansion accuracy against precise quadrature integration methods.
• Application to simulated chicken breeding data with body weight and leg deformity.

Main Results:

• Quadratic Taylor expansions demonstrated good accuracy in predicting responses for mixed trait data.
• The developed algorithms effectively handled selection processes with culling on scores.
• Taylor expansions provide a viable and less computationally intensive alternative to quadrature methods.
• Accuracy was validated using a realistic scenario involving chickens.

Conclusions:

• Taylor expansion-based algorithms are suitable for predicting BLUP selection responses with mixed continuous and discrete traits.
• These methods offer a practical solution for complex breeding schemes where computational resources are limited.
• The study provides valuable tools for enhancing genetic selection strategies in animal breeding.