MaizeGEP: A Maize Hybrids Dataset with Genotype, Phenotype, and Envirotype to Develop Genomic Selection Models.

Area of Science:

• Agricultural Science
• Genetics
• Data Science

Background:

• Precise phenotypic prediction requires integrating genotype and envirotype data.
• Existing datasets lack comprehensive environmental variables and broad trial locations.

Purpose of the Study:

• Introduce MaizeGEP, a novel dataset for genotype-envirotype to phenotype (GE2P) research.
• Facilitate advanced phenotypic prediction and variety selection in maize.

Main Methods:

• Developed MaizeGEP dataset with 260 maize varieties, 12,233 SNPs, 11 traits, 2382 year-county locations, and meteorological data.
• Employed mixture of experts (MoE) framework with GE2P algorithms for prediction.
• Utilized machine learning and deep learning models (SVM, LightGBM, MLP, DeepGS, DEM, Cropformer) for validation.

Main Results:

• MaizeGEP enables analysis of location clustering, population structure, and genome-wide associations.
• GE2P models and machine learning algorithms demonstrated effectiveness in phenotypic prediction.
• The dataset facilitates investigation of genotype-envirotype-phenotype relationships.

Conclusions:

• MaizeGEP is a valuable resource for understanding genotype-envirotype interactions and predicting cross-environmental performance.
• Encourages development of sophisticated GE2P models for improved plant breeding.
• Aids breeders in selecting and deploying maize varieties effectively across diverse environments.