Boosting predictabilities of agronomic traits in rice using bivariate genomic selection

Area of Science:

• Plant breeding and genetics
• Genomic selection
• Metabolomics

Background:

• Multivariate genomic selection (GS) models are understudied, limiting their application.
• Univariate GS models may not fully capture complex trait relationships.
• Integrating diverse omics data in GS requires efficient methodologies.

Purpose of the Study:

• To develop and evaluate a highly efficient bivariate (2D) genomic selection (GS) method.
• To compare the performance of the new 2D GS method against univariate (1D) GS models.
• To explore the utility of metabolomic traits in enhancing the predictability of traditional agronomic traits.

Main Methods:

• Development of a bivariate (2D) GS model incorporating the HAT (Haplotype-based Association Test) methodology.
• Application of the 2D BLUP-HAT GS model to a rice dataset analyzing traditional and metabolomic traits.
• Comparison of prediction accuracies between 2D BLUP-HAT GS and traditional 1D GS models.

Main Results:

• The 2D BLUP-HAT GS analysis demonstrated higher prediction accuracies for pairs of traits compared to individual trait analysis using 1D GS.
• Selected metabolites, when used as ancillary traits in the 2D BLUP-HAT GS method, significantly boosted the predictability of traditional traits.
• The computational efficiency of the 2D GS model was dramatically increased by avoiding conventional cross-validation through HAT methodology.

Conclusions:

• The developed 2D BLUP-HAT GS method offers significant advantages over univariate approaches for predicting multiple traits.
• Metabolomic data can be effectively integrated as ancillary traits to improve genomic prediction of important agronomic traits.
• This novel method enhances prediction accuracy and computational efficiency, paving the way for more effective breeding strategies.