Nitrogen nutrition index predicted by a crop model improves the genomic prediction of grain number for a bread wheat core collection

In plant breeding, one of the major challenges of genomic selection is to account for genotype-by-environment (G × E) interactions, and more specifically how varieties are adapted to various environments. Crop growth models (CGM) were developed to model the response of plants to environmental conditions. They can be used to characterize eco-physiological stresses in relation to crop growth and developmental stages, and thereby help to dissect G × E interactions. Our study aims at demonstrating how environment characterization using crop models can be integrated to improve both the understanding and the genomic predictions of G × E interactions. We evaluated the usefulness of using CGM to characterize environments by comparing basic and CGM-based stress indicators, to assess how much of the G × E interaction can be explained and whether gains in prediction accuracy can be made. We carried out a case study in wheat (Triticum aestivum) to model nitrogen stress in a CGM in 12 environments defined by year × location × nitrogen treatment. Interactions between 194 varieties of a core collection and these 12 different nitrogen conditions were examined by analyzing grain number. We showed that (i) CGM based indicators captured the G × E interactions better than basic indicators and that (ii) genomic predictions were slightly improved by modeling the genomic interaction with the crop model based characterization of nitrogen stress. A framework was proposed to integrate crop model environment characterization into genomic predictions. We describe how this characterization promises to improve the prediction accuracy of adaptation to environmental stresses