Identification of optimal prediction models using multi-omic data for selecting hybrid rice

Area of Science:

• Agricultural Science
• Genetics
• Plant Breeding

Background:

• Genomic prediction accelerates hybrid rice breeding cycles.
• Omics data (metabolomics, transcriptomics) are increasingly available for breeding value prediction.
• Optimizing prediction strategies is crucial for improving agronomically important traits.

Purpose of the Study:

• To determine the best prediction strategies for hybrid rice traits.
• To evaluate combinations of omics datasets and prediction methods.
• To identify optimal approaches for yield, grain weight, grain number, and tiller number prediction.

Main Methods:

• Comprehensive evaluation of all omics dataset combinations.
• Comparison of multiple prediction methods including BLUP, LASSO, SSVS, SVMs, and PLS.
• Analysis of genomic, metabolomic, and transcriptomic data for hybrid rice.

Main Results:

• Combined genomic and metabolomic data yielded superior prediction results compared to single omics or other combinations.
• Best Linear Unbiased Prediction (BLUP) demonstrated higher efficiency than other tested prediction methods.
• The study identified specific omics datasets and methods for improved trait predictability.

Conclusions:

• Integrating genomic and metabolomic data significantly enhances hybrid rice trait prediction.
• BLUP is the most effective prediction method for these traits.
• Findings provide practical guidelines to reduce costs and improve efficiency in hybrid rice breeding programs.