An R Package for Bayesian Analysis of Multi-environment and Multi-trait Multi-environment Data for Genome-Based...

Area of Science:

• Plant breeding
• Statistical genetics
• Bioinformatics

Background:

• Genomic selection (GS) is revolutionizing plant breeding by enabling genotype predictions without phenotyping.
• The success of GS heavily relies on statistical models, with a need for models tailored to various response variables (continuous, binary, etc.).
• Existing models primarily focus on univariate and continuous traits, lacking robust multivariate approaches for complex breeding data.

Purpose of the Study:

• To address the limitations of current multivariate statistical models in genomic prediction.
• To propose an improved Bayesian multi-trait and multi-environment (BMTME) R package for enhanced analysis of breeding data.
• To introduce computationally efficient Bayesian multi-output regressor stacking (BMORS) functions within the package.

Main Methods:

• Development and implementation of an improved BMTME R package.
• Integration of novel BMORS functions for efficient computational performance.
• Parameter estimation and prediction performance evaluation for multi-trait and multi-environment datasets.

Main Results:

• The enhanced BMTME package provides reliable, efficient, and user-friendly analysis of complex breeding data.
• BMORS functions offer significant computational efficiency compared to traditional BME() and BMTME() functions, especially for large datasets.
• The package facilitates parameter estimation and performance evaluation for multi-trait and multi-environment genomic predictions.

Conclusions:

• The improved BMTME R package, including BMORS functions, enhances the capabilities for genomic prediction in plant breeding.
• This tool supports more accurate and efficient prediction of candidate genotypes across multiple traits and environments.
• The development offers a valuable resource for researchers and breeders working with complex, high-dimensional breeding data.