关键词: 桃, 单果重, 可溶性固形物含量, 全基因组选择, 早期预测

Abstract: 【Objective】Fruit weight and soluble solid content (SSC) are two important quantitative traits in peach which are concerned by breeders and controlled by multiple minor genes. Therefore, it is difficult to perform early prediction by a single marker. As a novel genome-wide tool, genomic selection has been applied in fruit crops and expected to enhance breeding efficiency of those quantitative traits. However, its application effect in peach and influencing factors still need to be further explored.【Method】The objectives of this study were to assess the accuracy of prediction in nature and cross populations of peach for fruit weight and soluble solid content (SSC) by using genomic selection. In this study, the training population comprised 520 individuals were selected as materials. Using the genotypic data for 48,398 SNPs obtained from the resequencing results of above training population, a total of 11 genome-wide prediction models were built to select the optimum model for fruit weight and SSC. Then, we calculated the genomic breeding values of a small panel of nature population comprised 56 individuals and 29 cross populations comprising 1145 seedlings.【Result】The average sequencing data of each variety of the three groups was 1.95~3.52 Gb, and the sequencing depth was 5.29~10.79×. The sequencing data of the training natural population was algned with the reference genome, a total of 5,065,726 single nucleotide polymorphism (SNPs) were obtained. After removing the SNPs with high missing rate more than 20% and minor allele frequency lower than 0.05, a total of 48,398 SNPs on the genome were randomly selected for constructing whole-genome selection models for the training population. The model with the highest prediction accuracy for fruit weight is BayesA, and the model with the highest prediction accuracy for SSC is randomforest. Using the above two models, we found that the goodness of fit between the predicted breeding values and observed phenotype of fruit weight was 0.4767~0.6141, higher than that of SSC (0.3220~0.4329) in nature populations. And in cross populations, the prediction accuracy of fruit weight was 0.2319~0.4870, also showing higher than that of SSC (0.0200~0.2793).. The results also showed that the prediction model constructed by training natural populations was more accurate in predicting natural populations but of cross populations. Taking fruit weight as an example, we also found that only 17.78% of the seedlings need to be retained by genomic selection when targeted large fruit. Its efficiency is significantly higher than that of single and double marker selection. Finally, the effects of population dispersion, heritability and population structure on prediction accuracy also were discussed. The results indicated that prediction accuracy may be mixed and affected by a combination of several factors.【Conclusion】In this study, a suitable genomic selection model for peach fruit weight and SSC was screened and confirmed that the prediction efficiency of genomic selection was significantly higher than that of single marker selection. The results underline the potential of genomic prediction to accelerate breeding progress of these two quantitative traits in peach.

Key words: Peach, fruit weight, soluble solid contents, genomic selection, early prediction