Abstract:

Objective: This study aims to perform the genome selection (GS) for growth traits and the early selection of superior progeny in Castanopsis hystrix, which has great significance for rapid breeding of new superior varieties of C. hystrix. Method: In this study, 226 clones in the main distribution area and 479 progenies over 23 half-sib families were used as experimental materials. Genotyping datasets were obtained by high-throughput re-sequencing technology, and GS studies were conducted on the growth traits. The effects of 5 different GS models and 10 different numbers of SNPs on GS prediction accuracy were assessed using 5-fold cross-validation. Then, the genomic estimated breeding values (GEBV) of candidate populations were estimated based on the GS model, and early selection of superior progeny individuals was implemented by the Breggin multi-trait evaluation method. Result: The coefficient of variation of DBH trait in the training population was 22.73%, and greater than that of height trait (17.13%), and there was a significantly positive correlation between them (r=0.63, P<0.001). There were significant differences in both height and DBH among provenances (P<0.001). After re-sequencing and data quality control, 790 877 SNPs were obtained for each individual in the reference population and candidate population, and these SNPs were uniformly distributed in the C. hystrix genome. Based on the Genomic Best Linear Unbiased Prediction (GBLUP) model, the broad-sense heritability of height and DBH in the training population was 0.52 and 0.48, respectively, and the number of SNPs with different markers had little effect on the heritability estimation. Among the five GS models, Bayes B model had the highest GS prediction accuracy (0.21) for height, while Bayes ridge regression (BRR) model had the highest GS prediction accuracy (0.06) for DBH. The prediction accuracy of Bayes models was higher than that of GBLUP model, but the difference was not significant. For 10 different numbers of SNPs, the prediction accuracy of GS first increased during 0.5–5 K and then reached a stable stage. Bayes B model was used for height and Bayes RR model was used for DBH. The Brekin’s multi-trait evaluation method based on the GEBVs of these two traits was applied for the early selection of superior individuals in the candidate population. A total of 15 excellent progeny individuals were selected, and their mean GEBV values of height and DBH were 7.0% and 5.2% higher than those of the reference population, respectively. These superior offspring individuals were 4 438, 4 468, 4 407, 4 388, 4 052, 4 461, 4 390, 4 389, 4 410, 4 399, 4 460, 4 467, 4 212, 4 044, 4 459 and 4 020, mainly from two families of F5 and F29. Conclusion: In this study, a GS predicted model has been established, and the early selection of superior individuals has been carried out based on the GEBVs of the candidate populations, which lays the technical and material foundation for subsequent breeding of new superior varieties of C. hystrix.

Key words: Castanopsis hystrix, genomic selection, growth trait, early selection, SNP