Global Genotype by Environment Prediction Competition Reveals That Diverse Modeling Strategies Can Deliver...

Area of Science:

• Genomics
• Computational Biology
• Plant Science
• Animal Science

Background:

• Predicting phenotypes from genetic and environmental factors is a major challenge in modern biology.
• Improvements can lead to significant advancements in health, agriculture, and environmental management.
• The Genomes to Fields (G2F) initiative hosted a public Genotype by Environment (GxE) prediction competition in 2022-2023.

Purpose of the Study:

• To assess diverse predictive modeling strategies for genotype by environment interactions.
• To foster innovation by bringing together global participants from various disciplines.
• To leverage a comprehensive dataset of genomic, phenotypic, environmental, and management data.

Main Methods:

• Utilized a large, multi-year dataset encompassing genomic variation, phenotypes, weather, and field management data.
• Participants applied a wide range of modeling approaches, including machine learning, deep learning, quantitative genetics, and mechanistic models.
• The winning strategy combined machine learning (Random Forest, Ridge Regression, Least-squares) and traditional breeding tools, with separate models for environmental and genetic emphasis.

Main Results:

• The competition attracted a diverse global audience from academia, industry, and government.
• Various modeling techniques were employed, highlighting a breadth of approaches to GxE prediction.
• The winning approach integrated environmental and genetic factors using ensemble methods, while other high-performing teams used diverse strategies.

Conclusions:

• No single model or strategy demonstrated universal superiority for genotype by environment prediction.
• The diversity of successful methods underscores the complexity of GxE interactions and the value of interdisciplinary approaches.
• The competition provided valuable insights into modeling knowledge for GxE prediction using real-world data.