Files
Abstract
Genomic selection is a powerful tool for accelerating genetic gain in plant breeding by leveraging genome-wide markers. In this study, we evaluated genomic prediction accuracy across multiple scenarios using both simulated and real datasets. Our findings demonstrate that prediction accuracy was greater when phenotypes were strongly associated with genotypes and markers were selected based on GWAS significance rather than randomly. Traits controlled by fewer, large-effect QTLs consistently yielded higher accuracies, particularly at higher heritability levels. The inclusion of noise markers reduced accuracy, emphasizing the importance of precise and informed marker selection. Moreover, we showed that q-value-based marker selection effectively optimized prediction models, with intermediate q-value thresholds (0.1, 0.2) capturing nearly all true QTLs, minimizing the addition of non-informative markers. Validation with real data mirrored the simulation trends, and a QTL recovery analysis confirmed the reliability of this strategy for enhancing the prediction accuracy in genomic selection.