Hybrid kernels integrating genomic and multispectral data improve wheat genomic prediction accuracy

Area of Science:

• Plant breeding and genetics
• Agricultural science
• Bioinformatics

Background:

• Genomic selection (GS) enhances plant breeding efficiency but faces challenges in prediction accuracy.
• Factors like sample size and trait complexity impact GS model performance.
• Integrating genomic and phenomic data is a promising strategy to improve GS accuracy.

Purpose of the Study:

• To develop and evaluate a hybrid kernel modeling strategy for genomic selection.
• To improve the efficiency and accuracy of identifying superior genotypes in plant breeding.
• To leverage unmanned aerial vehicle (UAV)-derived phenomic data alongside genomic data.

Main Methods:

• Constructed hybrid kernels by combining genomic and phenomic data.
• Utilized multi-year winter wheat (Triticum aestivum L.) breeding data.
• Applied hybrid kernel modeling to UAV-derived phenomic and genomic data.

Main Results:

• Hybrid kernel modeling significantly enhanced GS prediction accuracy.
• Achieved average improvements of 17.52% (Pearson's correlation) and 30.36% (normalized RMSE).
• Demonstrated substantial gains in identifying top-performing lines (28.94% for top 10%, 16.73% for top 20%).

Conclusions:

• Integrating genomic and UAV-derived phenomic data via hybrid kernels is effective for improving GS.
• This approach offers a valuable and scalable tool for modern plant breeding.
• The method captures complementary variation, leading to more accurate genotype prediction.