Genomic models with genotype × environment interaction for predicting hybrid performance: an application in maize...

Area of Science:

• Agricultural Science
• Genetics
• Plant Breeding

Background:

• Predicting hybrid performance (HP) is crucial for agricultural breeding programs.
• Multi-environment trials are essential for selecting traits like yield and stability.
• Genotype-by-environment interaction (G×E) significantly influences plant performance across diverse conditions.

Purpose of the Study:

• To propose and evaluate a novel genomic statistical model incorporating G×E for predicting maize hybrid performance.
• To assess the model's ability to predict general and specific combining abilities across environments.
• To enhance the accuracy of predicting untested hybrid responses for key agronomic traits.

Main Methods:

• Developed a genomic statistical model that integrates G×E for predicting hybrid performance.
• Applied a cross-validation approach to evaluate two HP prediction models using extensive maize hybrid data.
• Utilized data from 2724 hybrids, 507 dent lines, and 24 flint lines across 58 environments over 12 years.

Main Results:

• Genomic models incorporating G×E demonstrated superior predictive ability compared to models without interaction.
• The improvement in predictive ability ranged from 12% to 22%, varying by trait.
• The model accurately predicted hybrid performance for traits including percent starch content, percent dry matter content, and silage yield.

Conclusions:

• Integrating genotype-by-environment interaction into genomic models significantly increases the accuracy of predicting untested maize hybrids.
• The proposed model offers a valuable tool for optimizing selection strategies in maize breeding.
• The model's framework is adaptable for predicting hybrid performance in other species with distinct parental pools.