Enviromic-based kernels may optimize resource allocation with multi-trait multi-environment genomic prediction for...

Area of Science:

• Agricultural Science
• Genetics
• Bioinformatics

Background:

• Genomic prediction accuracy relies on representative training sets, which are complex to establish, especially with multi-trait, multi-environment data.
• Envirotyping and correlated traits further complicate training set design for genomic prediction.

Purpose of the Study:

• To design optimized training sets for genomic prediction using multi-trait, multi-environment trial data.
• To evaluate strategies for increasing prediction accuracy and reducing phenotyping costs.

Main Methods:

• Two strategies were developed: one using genomic by environment by trait interaction (GET) and another including enviromic data (GWT).
• Genetic algorithms selected individuals based on prediction error variance to form optimized training sets, representing at least 98% of kernel variation.

Main Results:

• Combining genomic and enviromic data for optimized training sets improved response to selection per dollar by up to 145% compared to models without enviromic data.
• Prediction models incorporating genomic by environment (GxE) or enviromic data plus GxE showed superior prediction ability.

Conclusions:

• Genomic by enviromic by trait interaction kernels with genetic algorithms offer an efficient approach for designing optimized training sets.
• Significant improvements in genetic gains per dollar invested highlight the potential for effective resource allocation using this method.