Genomic-Enabled Prediction Based on Molecular Markers and Pedigree Using the Bayesian Linear Regression Package in R

Area of Science:

• Quantitative genetics
• Statistical genomics
• Bioinformatics

Background:

• Dense molecular markers enable genomic selection in plant and animal breeding.
• Current genomic selection models present computational and statistical challenges.
• Specialized software for genomic selection is not widely accessible or integrated into standard statistical packages.

Purpose of the Study:

• To introduce the R-package BLR for Bayesian Linear Regression (BLR) in genomic selection.
• To describe the statistical models implemented in the BLR package.
• To illustrate the practical application of BLR models using examples and address implementation challenges.

Main Methods:

• Implementation of Bayesian Ridge Regression and Bayesian LASSO within the BLR package.
• Joint analysis of marker genotypes and pedigree data in a unified framework.
• Description of model classes and their usage through practical examples.

Main Results:

• The BLR package provides a unified framework for advanced genomic selection models.
• It facilitates the joint use of marker and pedigree information.
• The package addresses practical challenges in applying genomic selection, including model choice and hyper-parameter selection.

Conclusions:

• The BLR package simplifies the implementation of genomic selection by integrating various Bayesian linear regression models.
• It offers a user-friendly solution for researchers and breeders facing computational and statistical hurdles.
• BLR enhances the application of genomic-enabled selection through accessible and integrated statistical procedures.