Integrating crop models, single nucleotide polymorphism, and climatic indices to develop genotype-environment...

Area of Science:

• Agricultural Science
• Genetics
• Computational Biology

Background:

• Genotype-environment interaction (G × E) models are crucial for digital breeding and predicting crop phenotypes.
• Genotype-specific parameters (GSPs) can bridge crop growth models and G × E interactions, simulating plant development.

Purpose of the Study:

• To integrate rice growth models, SNPs, and climate data for flowering time prediction.
• To investigate associations between GSPs and quantitative trait nucleotides (QTNs) using GWAS.
• To evaluate the impact of SNP-based GSPs and machine learning on prediction accuracy.

Main Methods:

• Utilized a dataset of 169 rice genotypes with 700K SNP markers and multi-environmental flowering data.
• Integrated three rice growth models (ORYZA, CERES-Rice, RiceGrow) with SNPs and climatic indices.
• Employed genome-wide association study (GWAS) to link GSPs with QTNs and machine learning (ML) for prediction refinement.

Main Results:

• Identified significant associations between GSPs and known rice flowering genes (e.g., DTH2, DTH3, OsCOL15).
• SNP-based GSPs in rice models led to decreased goodness of fit (increased RMSE) compared to traditional calibration.
• ML-modified predictions and multi-model ensembles achieved accuracy comparable to traditional methods.

Conclusions:

• GSPs offer genetic interpretability and potential for digital breeding applications in rice.
• Combining crop models with ML and climate data can improve phenotypic predictions.
• Findings support the advancement of molecular breeding strategies and accurate phenotypic prediction in rice.