A Genomic Bayesian Multi-trait and Multi-environment Model.

Area of Science:

• Quantitative genetics
• Statistical genomics
• Bioinformatics

Background:

• Traditional models for genotype × environment interaction (G × E) often focus on single traits.
• Comprehensive models incorporating correlated traits and trait × genotype × environment interaction (T × G × E) are needed for accurate whole-genome prediction (WGP).
• Existing methods lack the ability to simultaneously analyze multiple traits across multiple environments, limiting predictive power.

Purpose of the Study:

• To propose a novel Bayesian model for analyzing multiple traits and multiple environments for whole-genome prediction (WGP).
• To develop a computationally efficient Markov Chain Monte Carlo (MCMC) method for parameter estimation.
• To evaluate the proposed model's performance against existing methods using real datasets.

Main Methods:

• Developed a Bayesian model incorporating multiple traits and environments for WGP.
• Utilized Half-Cauchy priors for standard deviation terms and uniform priors for covariance matrix correlations.
• Implemented an efficient Gibbs sampling approach via MCMC for posterior distribution inference.

Main Results:

• The proposed Bayesian model demonstrated improved prediction accuracy compared to models with diagonal or standard variance-covariance structures when trait correlations were high (>0.5).
• The model's unstructured variance-covariance structure was particularly effective in scenarios with high trait correlations.
• An R-software package (BMTME) with optimized C++ routines was developed for efficient analysis.

Conclusions:

• The proposed Bayesian multi-trait, multi-environment model offers a significant advancement for whole-genome prediction.
• This method provides enhanced prediction accuracy, especially in complex scenarios involving correlated traits.
• The BMTME software package facilitates the practical application of this advanced statistical approach in genetic research.