| Sumario: | We explored the use of the eco-physiological crop model GECROS to identify markers for improved rice yield under well-watered (control) and water-deficit conditions. Eight model parameters were measured from the control in one season for 267 indica genotypes. The model accounted for 58% of yield variation among genotypes under control and 40% under water-deficit conditions. Using 213 randomly selected genotypes as training set, 90 SNP loci were identified using GWAS, explaining 42-77% of crop-model parameter variation. SNPs-based parameter values estimated from the additive loci effects were fed into the model. For the training set, the SNPs-based model accounted for 37% (control) and 29% (water-deficit) of yield variation, less than the 78% explained by a statistical genomic prediction (GP) model for the control treatment. Both models failed in predicting yields of the 54 testing genotypes. However, compared with the GP model, the SNPs-based crop model was advantageous when simulating yields under either control or water-stress conditions in an independent season. Crop-model sensitivity analysis ranked the SNP loci for their relative importance in accounting for yield variation, and the rank differed greatly between control and water-deficit environments. Crop models have a potential to use single-environment information for predicting phenotypes under different environments.
|
|---|