Cultivar differentiation of fritillary using machine learning model-driven untargeted metabolomics

Fritillaria cirrhosa (F. cirrhosa), a medicinal and edible herbs, is highly regarded for its remarkable efficacy in relieving cough and eliminating phlegm. However, surging demand and wild resource shortages have caused widespread adulteration of F. thunbergii with F. ussuriensis. Therefore, conducting a comparative study of these types of fritillaries may be helpful for the quality control and identification of the F. cirrhosa medicinal material. This study aims to conduct a comparative study of the chemical components of three types of fritillaries by integrating plant metabolomics combined with machine learning, from the perspectives of component analysis, differential identification, and quantitative analysis.The chemical components of fritillaries were comprehensively characterized using UHPLC-Q-Exactive Orbitrap MS technology, and a total of 220 chemical components were identified. Combined with chemometrics methods,untargeted metabolomics and six machine learning models, eight differential components were further selected, and their contents were determined. Among them, sipeimine, edpetiline, solanidine, ebeiedinone, and peimisine may serve as differential components for distinguishing F. cirrhosa from F. thunbergii and F. ussuriensis, providing chemical basis for the authenticity identification and quality control of F. cirrhosa.