Genomic selection (GS) is a promising tool to accelerate genetic gain for complex traits. In this study, we evaluated the potential of GS for the improvement of seven lodging-related traits in double-cropping rice in Southern China using 438 rice accessions. The traits examined included the length and bending resistance of the third and fourth internodes (IL3, IL4, BR3, BR4), plant height (PH), and the ratio of internode length to plant height (IL3/PH, IL4/PH). Significant phenotypic differences were observed for all traits between the two seasons. In comparisons of cross-validation and independent prediction, GBLUP and BayesLASSO outperformed LightGBM across all traits in both seasons. Across all evaluated traits, prediction accuracies (Pearson’s r) ranged from 0.33 to 0.78 in cross-validation and from 0.28 to 0.75 in independent prediction using the GBLUP model. Bending resistance exhibited lower prediction accuracy due to its lower genomic heritability. Correlation analysis revealed that plant height was not significantly correlated with culm bending resistance, suggesting that these traits are genetically independent. We utilized GBLUP models trained on our experimental data to predict the genomic estimated breeding values (GEBVs) of the 3000 Rice Genomes Project (3kRG) dataset. The results demonstrated that GS can efficiently enrich the proportion of highly lodging-resistant accessions, increasing it from 31.40% in the base 3kRG population to a maximum of 83.00% among the top 200 selected individuals. Furthermore, indirect selection for traits with higher heritability, such as IL and IL/PH, was more effective at screening highly lodging-resistant cultivars than direct selection for BR. Our research demonstrates the feasibility of applying genomic selection for the breeding of lodging-resistant varieties in double-cropping rice and provides a foundation for further applications.
Lu et al. (Wed,) studied this question.