A singular value decomposition Bayesian multiple-trait and multiple-environment genomic model.

Area of Science:

• Quantitative Genetics
• Genomic Selection
• Plant and Animal Breeding

Background:

• Genomic-assisted breeding aims to improve multiple traits simultaneously.
• Current multi-trait, multi-environment models present implementation challenges for breeders.
• Efficient algorithms are needed for complex genomic analyses in breeding programs.

Purpose of the Study:

• To propose a novel, computationally efficient four-stage analysis for multiple-trait genomic data.
• To evaluate the performance of the proposed method against existing Bayesian multi-trait and multi-environment models.
• To demonstrate the applicability of the new method using standard single-trait genomic prediction software.

Main Methods:

• A four-stage analysis framework is introduced for multiple-trait data.
• Stage 1: Singular Value Decomposition (SVD) on the trait response matrix.
• Stages 2-4: Multiple-trait analysis on transformed responses, followed by back-transformation for parameter estimation and prediction.

Main Results:

• The proposed SVD-based method achieved comparable parameter estimation and prediction accuracy to the Bayesian multi-trait and multi-environment model (BMTME).
• The SVD approach demonstrated similar performance metrics to the established BMTME model.
• The method's results were validated against the BMTME model in terms of accuracy and parameter estimation.

Conclusions:

• The proposed four-stage SVD method provides an effective alternative for multiple-trait genomic analysis.
• This approach simplifies implementation by leveraging existing single-trait genomic prediction software.
• The method offers a computationally efficient algorithm, making advanced genomic selection more accessible to breeders.