Optimizing genomic selection models for wheat breeding under contrasting water regimes in a mediterranean environment
Background Bread wheat (Triticum aestivum L.) is a vital global crop, supplying 20% of the protein in the human diet. Improving its productivity and resilience, particularly under water-limited conditions, is a major breeding priority. Genomic selection offers a promising approach to accelerate gene...
| Autores: | , , |
|---|---|
| Tipo de recurso: | artículo |
| Estado: | Versión publicada |
| Fecha de publicación: | 2025 |
| País: | España |
| Institución: | Varias* (Consorci de Biblioteques Universitáries de Catalunya, Centre de Serveis Científics i Acadèmics de Catalunya) |
| Repositorio: | Recercat. Dipósit de la Recerca de Catalunya |
| OAI Identifier: | oai:recercat.cat:10459.1/469693 |
| Acceso en línea: | https://doi.org/10.1186/s13007-025-01467-5 https://hdl.handle.net/10459.1/469693 |
| Access Level: | acceso abierto |
| Palabra clave: | Genomic selection Wheat breeding Water stress Machine learning models |
| Sumario: | Background Bread wheat (Triticum aestivum L.) is a vital global crop, supplying 20% of the protein in the human diet. Improving its productivity and resilience, particularly under water-limited conditions, is a major breeding priority. Genomic selection offers a promising approach to accelerate genetic gains by predicting complex traits using genome-wide marker data. This study evaluated the performance of various genomic selection (GS) models in predicting key agronomic traits under contrasting well-watered (WW) and water-stressed (WS) conditions, with the goal of enhancing drought adaptation in wheat breeding programs. Results A panel of 179 wheat lines was evaluated for grain yield, yield components, and grain protein content. Models were trained on data from well-watered and water-stressed regimes independently, as well as on combined data from both conditions. Predictive approaches included linear models (Ridge Regression Best Linear Unbiased Prediction and Bayesian methods), semi-parametric models (Reproducing Kernel Hilbert Space Regression), and machine learning algorithms (Random Forest, Support Vector Machine, and Extreme Gradient Boosting). Ridge regression consistently delivered strong performance across all traits and conditions, with mean rMG of 0.70 (water-stressed), 0.64 (well-watered), and 0.65 (combined). Machine learning models, especially Random Forest and Extreme Gradient Boosting, performed competitively for complex traits such as grain yield and thousand kernel weight. Random Forest achieved a rMG of 0.81 for grain yield and 0.73 for thousand kernel weight under well-watered conditions. Trait stability was observed under well-watered conditions for thousand kernel weight and plant height, supported by moderate heritability estimates (0.69–0.74). Cross-validation comparisons showed consistent model performance across validation schemes, with leave-one-out cross-validation offering slightly improved accuracy for select traits and models. Notably, models trained under water-stressed conditions generalized better when tested on well-watered data than the reverse, highlighting the value of diverse training environments. Conclusions Genomic selection models, particularly ridge regression and machine learning approaches, demonstrated reliable predictive performance across environments and traits. Incorporating data from multiple environmental conditions improves prediction accuracy and supports the development of drought-resilient wheat lines. These results reinforce the utility of genomic selection in modern wheat breeding programs for enhancing both productivity and stress tolerance. |
|---|