The study integrated three rice growth models (ORYZA, CERES-Rice, and RiceGrow), genome-wide association studies (GWAS), and climate indices. Machine learning algorithms were then employed to correct prediction errors. The results suggest this hybrid modeling approach could significantly enhance the accuracy of flowering time predictions and support genotype selection in rice breeding programs.
Accurately predicting crop traits such as flowering time is essential for modern rice breeding, especially under changing environmental conditions. Understanding how genotype (G), environment (E), and their interaction (G×E) influence crop phenotypes is a core challenge in breeding. While process-based crop models simulate plant development based on environmental and physiological processes, they often overlook complex genetic contributions. Recent efforts have focused on using genomic data—especially single nucleotide polymorphisms (SNPs)—to predict genotype-specific parameters (GSPs), bridging genotype and phenotype. Yet, these predictions often fall short in accuracy due to the models’ limited capacity to capture nonlinear G×E effects, especially under stress conditions. To overcome these challenges, integrating machine learning and climate variables into crop models offers a promising route for improving prediction precision and interpretability.
A study (DOI: 10.1016/j.plaphe.2025.100007) published in Plant Phenomics on 25 February 2025 by Liang Tang’s team, Nanjing Agricultural University, offers a novel pathway to more interpretable, transferable, and robust prediction systems, holding great promise for precision agriculture and molecular breeding.
In this study, researchers employed a integrative modeling approach that combined process-based rice growth models, genome-wide association studies (GWAS), SNP-based genomic prediction, and machine learning algorithms to predict rice flowering time across diverse genotypes and environments. First, genotype-specific parameters (GSPs) were estimated for three crop models (ORYZA, CERES-Rice, and RiceGrow), revealing both normal and bimodal distributions among key photoperiod and temperature sensitivity traits. Significant variation was observed in GSPs such as PhotoDCERES and IntriERiceGrow, with coefficients of variation exceeding 60% in several cases. Correlation analysis indicated substantial consistency among photoperiod-related GSPs across models. The accuracy of flowering time predictions using these fitted GSPs was high, with root mean square errors (RMSEs) of 10.11–21.25 days and Pearson correlation coefficients reaching up to 0.94. GWAS analysis identified hundreds of quantitative trait nucleotides (QTNs) linked to specific GSPs, including markers near known flowering time genes such as DTH2, DTH3, DTH7, and OsCOL15. Next, SNP-based predictions of GSPs were conducted using ridge regression and rr-BLUP; ridge regression outperformed rr-BLUP, particularly in the test set. Although SNP-predicted GSPs initially resulted in reduced prediction accuracy, integration with machine learning models—especially XGBoost—significantly restored performance in a second modeling stage. Among climate indices considered, growing degree days (GDD) at 100 days after sowing emerged as the most critical factor across all models. The multi-model ensemble (MME) approach showed robust performance, often matching the accuracy of the best individual model. These findings highlight the value of combining mechanistic modeling, genomic data, and climate-informed machine learning to improve the accuracy and interpretability of phenotype predictions in rice breeding.
This integrative G×E modeling strategy holds great promise for accelerating molecular breeding in rice. By improving the prediction of complex traits like flowering time, breeders can better select genotypes tailored for specific climates, reducing development cycles and enhancing crop resilience. The study also provides a scalable blueprint for other crops, especially those facing unpredictable weather and environmental stress.
###
References
DOI
10.1016/j.plaphe.2025.100007
Original Source URL
https://doi.org/10.1016/j.plaphe.2025.100007
Funding information
This study was supported by the National Key Research and Development Program of China (2022YFD2001001), the Jiangsu Independent Innovation Fund Project of Agricultural Science and Technology [CX(21)1006], the Jiangsu Collaborative Innovation Center for Modern Crop Production (JCICMCP), and the 111 Project.
About Plant Phenomics
Science Partner Journal Plant Phenomics is an online-only Open Access journal published in affiliation with the State Key Laboratory of Crop Genetics & Germplasm Enhancement, Nanjing Agricultural University (NAU) and distributed by the American Association for the Advancement of Science (AAAS). Like all partners participating in the Science Partner Journal program, Plant Phenomics is editorially independent from the Science family of journals. Editorial decisions and scientific activities pursued by the journal's Editorial Board are made independently, based on scientific merit and adhering to the highest standards for accurate and ethical promotion of science. These decisions and activities are in no way influenced by the financial support of NAU, NAU administration, or any other institutions and sponsors. The Editorial Board is solely responsible for all content published in the journal. To learn more about the Science Partner Journal program, visit the SPJ program homepage.