GWAS-driven gene mining and genomic prediction of ornamental traits in flowering trees: A case study of Camellia japonica

Abstract

Floral traits largely determine the ornamental value of horticultural plants, while the long juvenile period of woody plants hinders the progress of the floral pattern and color breeding. The population genetic variations in the floral characters of cultivated camellias are far less well understood and applied. Here, we investigated genetic architecture and genome prediction of the floral pattern and color in Camelia japonica. Seven anthocyanins were identified in 200 camellia cultivars using an ultra-high performance liquid chromatography-mass spectrometry (UPLC-MS) approach. The content and proportional changes in Cy3G and Cy3GEpC were identified as the main cause of the color change. A total of 2 072 667 SNPs were identified, the population structure analysis revealed frequent gene infiltration among the cultivars. A genome-wide association study (GWAS) and the transcriptome analysis identified 163 and 46 shared genes significantly associated with the floral color and pattern, respectively. Furthermore, Support Vector Machine (SVM) regression with linear kernel and the top 1000 and 10 000 GWAS associated markers achieved the highest prediction accuracy for a petal number of 94 %, and anthocyanin content of 95 %. Our study provides novel insight into the genetic basis of floral characters and confirms the feasibility of using machine learning and GWAS markers to predict floral traits, which will accelerate the ornamental molecular breeding of C. japonica.