Origin traceability of Yimucao (Chinese motherwort) in China using stable isotopes and extracts assisted by machine learning techniques

Abstract

Leonurus japonicus Houtt. is a medicine food homology plant that is widely farmed in China. In traditional Chinese medicine, the aerial part of L. japonicus (Chinese motherwort) is named Yimucao and has medicinal uses. Yimucao in the seedling stage can be eaten as a wild vegetable and incorporated into one’s everyday diet. The quality of Yimucao is often associated with its production origins, and the geographical authenticity of Yimucao is important for ensuring its clinical efficacy. A combined strategy based on the analysis of stable isotopes (δ13C, δ15N, δ2H, and δ18O), elemental content (%C and %N), and extracts (aqueous and ethanol extracts) was conducted to trace the geographical origin of Yimucao in China. Here, eight variables of 63 Yimucao samples collected from eight provinces were examined, and notable distinctions were observed on the provincial scale and regional scale (P < 0.05). Principal component analysis, orthogonal partial least square–discriminant analysis, and four machine learning methods (random forest, adaptive boosting, support vector machine, and neural network) were applied for geographical classification. We found that the random forest model was the most optimal classifier with a remarkable prediction accuracy reaching 98.4%. Among the eight differentiation markers analyzed, δ15N, δ18O, and δ2H were the most potent indicators. The correlation analysis between eight variables and environmental factors indicated that latitude, sunshine duration, and relative humidity were responsible for the majority of the differences in the production areas. This study demonstrated that comprehensive analysis of stable isotopes and extracts assisted by machine learning algorithms is a powerful method for determining the geographical origins of Yimucao in China.