Optimal Designs for Genomic Selection in Hybrid Crops.

Area of Science:

• Genomics
• Biotechnology
• Plant Breeding

Background:

• Genomic selection leverages whole-genome genotype-phenotype relationships for crop improvement.
• Current methods for genomic prediction of hybrid performance can be further optimized.

Purpose of the Study:

• To optimize genomic prediction of hybrid performance using design-thinking and data-mining techniques.
• To develop efficient training set designs for genomic prediction models.

Main Methods:

• Phenotyped maize, wheat, and rice hybrids for key agronomic traits.
• Utilized 10,296,310 SNPs from parental inbreds.
• Employed clustering, network analysis, and genetic mating schemes to design training samples.

Main Results:

• Optimized training set designs significantly outperformed random sampling and previous methods.
• Effective genomic prediction models were established using only 2%-13% of the total hybrid set.
• Validated approaches across maize, wheat, and rice breeding systems.

Conclusions:

• Design-thinking and data-mining offer a powerful strategy for optimizing genomic prediction of hybrid performance.
• Efficient training set selection enables cost-effective exploration of vast genetic combinations in crop breeding.
• This approach enhances the efficiency of genomic selection in diverse hybrid breeding programs.