A stochastic simulation study on validation of an approximate multitrait model using preadjusted data for prediction...

Area of Science:

• Animal Breeding and Genetics
• Quantitative Genetics
• Dairy Cattle Production

Background:

• Accurate prediction of breeding values is crucial for genetic improvement in dairy cattle.
• Selection goals often involve multiple traits, necessitating sophisticated evaluation models.
• Previous studies have explored various genetic models, but direct comparisons in large populations are essential.

Purpose of the Study:

• To compare the effectiveness of three different models for predicting breeding values in a large dairy cattle population.
• To evaluate the genetic response achieved by univariate, approximate multitrait, and linear multitrait models.
• To determine the optimal model for genetic gain considering both accuracy and practicality.

Main Methods:

• Stochastic simulation of a 100,000-cow dairy cattle population over 35 years.
• Initial selection phase using univariate and trivariate models.
• Comparison of four scenarios: univariate, trivariate, approximate multitrait, and linear multitrait models.
• Assessment based on the regression coefficient of true genetic values on year.

Main Results:

• The linear multitrait model yielded the highest genetic gain (3.073 +/- 0.069 economic units/year).
• The approximate multitrait model (2.819 +/- 0.047) significantly outperformed the univariate approach (2.672 +/- 0.060).
• Linear multitrait models are limited in the number of traits they can handle compared to approximate models.

Conclusions:

• The linear multitrait model provides the most accurate prediction of breeding values and maximizes genetic response.
• The approximate multitrait model is a viable and practical option for countries developing total merit indices that incorporate numerous traits.
• Future developments should focus on enhancing approximate models to handle complex genetic correlations effectively.