Discrimination of the geographic origins and varieties of wine grapes using high-throughput sequencing assisted by a random forest model

Abstract

The geographic origin and variety of wine grapes play a vital role in shaping wine characteristics. To determine the geographic origins and varieties of wine grapes, the compositions of the microbial communities on the surface of four varietal grapes from five regions in Xinjiang, China were investigated using high-throughput sequencing. Furthermore, a random forest (RF) classifier was introduced to distinguish the geographic origins and varieties of the wine grapes. The results show that the construction of the microbial communities on the surfaces of the wine grapes is driven more by the geographic origin than the variety. The fungal RF classifier shows better accuracy than the bacterial RF classifier in determining the geographic origins and varieties of wine grapes. The fungal RF classifier exhibits a strong ability to authenticate the geographic origins of Cabernet Sauvignon based on the composition of fungi on the surface of the grapes with an accuracy of 93.33%. The RF classifier exhibits the highest accuracy (66.67%) in distinguishing the wine grape varieties. Thus, this study confirmed the potential of the HTS technique coupled with the random forest algorithm for identifying the geographic origin and variety of wine grapes.