Genomic predictions in diploid Actinidia chinensis (kiwifruit)

Kiwifruit are dioecious, long-lived, perennial, woody vines. The consequences for kiwifruit breeding include lengthy breeding cycles, large planting areas and selection of male parents by progeny tests, which are expensive and time-consuming. Also, most economically important traits, such as yield, fruit quality and disease resistance, are complex because of their polygenic (quantitative) nature. Genomic selection in kiwifruit has been made possible because of the development of restriction enzymebased genotyping-by-sequencing (GBS), together with new statistical methods for analysing genotypic data. However, information on model robustness and guidelines for model selection are lacking in kiwifruit. In this study, a training population consisted of 88 plants, including four of the parents and 84 seedlings, which were randomly selected from a diploid Actinidia chinensis breeding population. Of these, 78 were used as a training set and 10 were used as a validation set. Random tagging GBS (rtGBS) libraries were constructed using the restriction enzyme BamHI and sequenced on Illumina (R) HiSeq2000 platform in Single-end mode. A total of 8372 SNP markers were obtained on all 88 individuals. Three statistical models, RR-BLUP (ridge regression best linear unbiased predictor), GBLUP (genomic best linear unbiased predictor), and LASSO (least absolute shrinkage and selection operator), were used to estimate GEBVs (genomic estimated breeding values) for nine traits. The correlations between predicted GEBVs and observed phenotypic data or breeding values estimated from phenotypic data of the validation set were calculated to measure the reliabilities of the predicted GEBVs. The correlations ranged from 0.48 to 0.91 for red intensity, fruit weight, yield, and fruit firmness at 60, 90 and 120 days of storage, whereas negative correlations were obtained for dry matter, soluble solids content and Pseudamanas syringae pv. actinldlae (Psa) resistance. Possible reasons for the negative correlations were discussed. Our results showed that the GBLUP model can be used to predict GEBVs in kiwifruit breeding.