Increasing Genomic-Enabled Prediction Accuracy by Modeling Genotype × Environment Interactions in Kansas Wheat

Area of Science:

• Plant breeding
• Quantitative genetics
• Agricultural science

Background:

• Wheat breeding programs rely on multi-location, multi-year trials for accurate yield and stability assessment.
• Early generation selections have limited data, hindering understanding of genotype × environment (G × E) interactions.
• Genomic selection (GS) offers a way to integrate G × E effects for improved prediction.

Purpose of the Study:

• To evaluate the increased breeding efficiency using genomic selection (GS) reaction norm models in hard red winter wheat breeding.
• To assess the predictive accuracy of GS models incorporating various effects (line, environment, genotype, G × E) for yield performance.

Main Methods:

• Utilized reaction norm models with cross-validation schemes to analyze wheat yield data.
• Compared GS models with different covariate inclusions: L + E, L + E + G, and L + E + G + (G × E).
• Evaluated prediction accuracy for untested environments, sites, and forward prediction (future years).

Main Results:

• GS reaction norm models demonstrated high accuracy (>0.4) in predicting yield for untested environments/sites.
• The GS model including genotype × environment (G × E) effects (L + E + G + (G × E)) showed the highest prediction ability ( = 0.54) for incomplete trials.
• Forward prediction accuracy was relatively low ( = 0.171), highlighting challenges in predicting future years.

Conclusions:

• Genomic selection reaction norm models significantly enhance wheat breeding efficiency by improving prediction accuracy across diverse environments.
• The inclusion of genotype × environment interactions is crucial for accurate yield prediction in multi-environment trials.
• While GS improves prediction for untested locations, predicting future year performance remains a challenge.