GWAS-assisted genomic selection for achieving high-precision early sex prediction in Populus deltoides

Poplar (Populus spp.) breeding programs increasingly prioritize the development of male varieties due to the environmental issues caused by female catkins. Therefore, there is an urgent need for a reliable technique for early sex identification in Populus. Genomic selection (GS), as an efficient predictive method, offers a promising solution for early sex identification in poplar. In this study, we conducted genomic prediction for sex in Populus deltoides. Using five full-sib families of P. deltoides as the reference population, we identified 801 sex-associated loci through GWAS and precisely localized the sex determination region at the telomeric end of chromosome 19. We evaluated 14 GS statistical models using fivefold cross-validation under six marker densities. The results showed significant differences in prediction accuracy (PA) among different statistical models, ranging from 0.19 to 0.79, with the gradient boosting decision tree exhibiting the highest accuracy and stability. Notably, single nucleotide polymorphisms selected through GWAS significantly improved PA compared to random markers, achieving a corrected accuracy of 0.999. Using the optimal model and markers, we predicted the sex of 505 progenies from 27 full-sib families, with over 90% of the predictions being accurate. Overall, this study achieved high-accuracy sex prediction in P. deltoides through genome prediction, providing a novel and efficient method for poplar sex identification.