Genetic Parameter and Hyper-Parameter Estimation Underlie Nitrogen Use Efficiency in Bread Wheat.

Area of Science:

• Plant breeding
• Genetics
• Agricultural science

Background:

• Phenotyping complex traits like nitrogen use efficiency (NUE) in wheat is costly and time-consuming compared to genotyping.
• Predicting phenotypes from marker information is a key strategy in breeding programs.
• Accurate estimation of genetic parameters and hyperparameters is crucial for reliable genomic selection (GS) models.

Purpose of the Study:

• To evaluate the performance of an ensemble-learning-based model (STACK) for estimating genetic parameters and hyperparameters in genomic selection.
• To address the challenge of ensuring reliability and accuracy in predicted breeding values (BVs).

Main Methods:

• Utilized an ensemble-learning-based model named STACK.
• Investigated the impact of various genetic parameters (e.g., population structure, marker density) and hyperparameters (e.g., panel size, marker count) on model performance.
• Confirmed bias-variance tradeoff and adaptive prediction error.

Main Results:

• The STACK model demonstrated the highest performance in estimating genetic parameters and hyperparameters within a genomic selection model.
• The study validated the bias-variance tradeoff and adaptive prediction error principles for the STACK model.
• STACK outperformed other models in the context of estimating these critical factors.

Conclusions:

• Ensemble learning, specifically the STACK model, offers a superior approach for estimating genetic parameters and hyperparameters in genomic selection.
• Accurate estimation of these factors is vital for improving the reliability and accuracy of predicted breeding values in wheat breeding programs.
• The findings contribute to more efficient and accurate prediction of complex traits like NUE in crop improvement.