Predicting quantitative traits with regression models for dense molecular markers and pedigree

Area of Science:

• Quantitative genetics
• Genomics
• Statistical genetics

Background:

• Genomewide dense markers offer new possibilities for animal and plant breeding.
• Predicting genetic values for complex traits requires integrating markers, pedigrees, and phenotypes.
• Standard regression models may require shrinkage for numerous markers.

Purpose of the Study:

• To adapt the Bayesian least absolute shrinkage and selection operator (LASSO) for joint analysis of markers, pedigrees, and covariates.
• To evaluate the performance of this adapted Bayesian LASSO model in predicting genetic values.
• To assess the impact of marker inclusion on predictive ability.

Main Methods:

• Developed a regression model incorporating markers, pedigrees, and covariates using Bayesian LASSO.
• Investigated connections between Bayesian LASSO and other marker-based regression models.
• Assessed the sensitivity of the model to prior distribution choices via simulation.
• Applied the model to wheat and mouse population data, using cross-validation for evaluation.

Main Results:

• The adapted Bayesian LASSO model successfully integrated markers, pedigrees, and covariates.
• Inclusion of markers significantly improved the predictive ability of the models.
• The model demonstrated robustness and effectiveness in genetic value prediction.

Conclusions:

• The proposed Bayesian LASSO approach enhances the prediction of genetic values for complex traits.
• Integrating genomewide markers alongside traditional data improves breeding program outcomes.
• Freely available R software facilitates the application of this advanced statistical method.