Application of stable isotopes with machine learning techniques for identifying Aconiti Lateralis Radix Praeparata (Fuzi) geographical origins

Abstract

Authentication of geographical traceability is essential to the safety of traditional Chinese medicines (TCMs). Aconiti Lateralis Radix Praeparata (Chinese: Fuzi), a well-known TCMs, the therapeutic efficacy and toxicity are clearly correlated with its geographical origin, and a rapid and effective method of origin tracing must be established. Herein, an elemental analysis-isotope ratio mass spectrometry (EA-IRMS) combined with a machine learning model is constructed for the origin traceability of Fuzi. We analyzed stable isotope information (δ13C‰, δ15N‰, δ2H‰, δ18O‰, C% and N%) for a total of 243 samples from eight major producing areas. Classification models were built using seven machine learning techniques, including random forest (RF), support vector machine (SVM), AdaBoost, etc., to investigate the feasibility of stable isotopes in Fuzi geographic identification. Results show that stable isotopes exhibit an overwhelming origin discrimination advantage. Among them, the RF model showed the best recognition performance with 94.2 % accuracy, 95.45 % sensitivity, and 99.15 % specificity in the independent test set, respectively. The results show that stable isotopes combined with machine learning techniques have great potential for the origin traceability of TCMs.