learnMET: an R package to apply machine learning methods for genomic prediction using multi-environment trial data.

Area of Science:

• Agricultural Science
• Bioinformatics
• Computational Biology

Background:

• Multi-environment trials (MET) are crucial for assessing genotype performance across diverse conditions.
• Integrating genomic and environmental data can enhance the accuracy of plant breeding predictions.
• Existing tools may lack flexibility in handling diverse environmental data and machine learning approaches.

Purpose of the Study:

• To introduce learnMET, an R-package designed for comprehensive analysis of MET breeding data.
• To provide a flexible framework for incorporating genomic, climate, and soil data into machine learning models.
• To facilitate advanced genomic prediction and model evaluation for plant breeders.

Main Methods:

• Development of an R-package, learnMET, offering a unified framework for MET data analysis.
• Integration of genomic data with environmental variables, including field and NASA-sourced weather data.
• Implementation of various machine learning algorithms (e.g., gradient boosting, random forests, MLPs) for genomic prediction.
• Inclusion of diverse cross-validation strategies tailored for MET experimental designs.

Main Results:

• learnMET enables the combination of diverse data types (genomic, environmental) for sophisticated analyses.
• The package supports flexible aggregation of daily weather data using phenological or naive approaches.
• Multiple machine learning models are available for genomic prediction, with robust evaluation schemes.
• The package is user-friendly, aiding plant breeders in practical application.

Conclusions:

• learnMET provides a powerful and flexible R-package for machine learning-based genomic prediction in plant breeding.
• The integration of environmental data enhances the predictive power of breeding programs.
• The package democratizes access to advanced analytical methods for multi-environment trial data.