1H NMR-based metabolomic approach combined with machine learning algorithm to distinguish the geographic origin of huajiao (Zanthoxylum bungeanum Maxim.)

Abstract

Chinese prickly ash, or huajiao (Zanthoxylum bungeanum Maxim), is the widely favored species of this spice by both the agricultural industry and the gourmet market. The geographic origin of the spice can add value, creating challenges for quality control and brand protection. In this study, 160 samples from four main production regions were analyzed by 1H NMR spectroscopy for the first time. Data analysis was performed based on 32 non-polar metabolites. Several machine learning algorithms were tested to build classification models. The best modeling results were obtained by using the non-linear discriminant random forest model, achieving an overall accuracy of 100% for the training set, 95.7% for the test set, and 87.5% for the blind dataset. The main marker compounds responsible for distinguishing these four origins were linalool, linalyl acetate, nonanal, and ocimene. This study provides a method to use 1H NMR combined with chemometrics to determine the origin of the huajiao.