learnMET

learnMET performs genomic prediction by integrating genomic data with climate and soil variables to model genotype-by-environment interactions in multi-environment trial (MET) breeding datasets for plant breeding.

Key Features:

• Integration of Genomic and Environmental Data: Combines genomic information with environmental variables including climate and soil characteristics, using weather data from field stations or global meteorological datasets sourced from a NASA database.
• Data Aggregation Methods: Aggregates daily weather data over specified periods using naive approaches (e.g., nonoverlapping 10-day windows) or phenological approaches for temporal environmental summarization.
• Machine Learning Models for Genomic Prediction: Implements gradient-boosted decision trees, random forests, stacked ensemble models, and multilayer perceptrons for predicting phenotypic outcomes.
• Cross-Validation Schemes: Provides an array of cross-validation schemes that simulate real-world MET scenarios for robust model evaluation and validation.

Scientific Applications:

• Genomic prediction across environments: Predicts phenotypic outcomes across diverse environments by combining genetic and environmental information.
• Genotype-by-environment interaction analysis: Assesses how genotypes respond to varying climate and soil conditions within MET datasets.
• Selection of resilient genotypes: Supports selection of genotypes with desirable traits under climate variability and other environmental stresses.

Methodology:

Integrates genomic data with environmental datasets (weather from field stations or a NASA database), aggregates daily weather using nonoverlapping 10-day windows or phenological approaches, applies machine learning models including gradient-boosted decision trees, random forests, stacked ensemble models, and multilayer perceptrons, and evaluates models using MET-oriented cross-validation schemes.