Heterosis prediction system based on non-additive genomic prediction models in cucumber (Cucumis sativus L.)

Abstract

The heterosis prediction has an essential role in hybrid breeding. The prediction of heterosis is highly dependable upon the selection of an efficient method, however, fewer methods have been used till date which lack or unable to explain several important aspects in heterosis prediction. Thus, in our experiment, we tried to explore a new heterosis prediction system based on genomic prediction, in cucumber breeding. Two models, SNP-BLUP and GBLUP were used for genomic prediction model training. The additive, and dominance components were estimated by SNP-BLUP; General combining ability (GCA), and special combining ability (SCA) components were estimated by GBLUP. We explored the correlation of these four model components, the genetics distances with heterosis. The our results show that both the dominance and SCA have stronger correlations with heterosis (r=0.36∼0.73). While genetic distance, additive model component, and GCA have relatively weak or no significant correlations with heterosis, which may not be suitable for the prediction of heterosis. Therefore, the GD model, Dominance model, and SCA model based on genetic distance, dominance, and SCA predictors were established, respectively. Compared with the GD model, the predictive abilities of the Dominance model and the SCA model were improved by 0.15∼0.74 and 0.15∼0.59, respectively. In conclusion, we have developed an improved heterosis prediction system, which might offer valuable information for cucumber breeding and also provide a reference for the prediction of heterosis in other horticulture crops.