Chemometric strategies for binary discrimination and one-class classification of medicinal parts of Angelicae Sinensis Radix by fusing chemical profiles with anticoagulant activity

Angelicae Sinensis Radix (AS), an herb and functional food, has different medicinal parts with distinct clinical efficacies and market values. This study aimed to discriminate AS and its different parts (head, body, and tail) by integrating chemical profiles with anticoagulant activity. Fourier-transform infrared (FT-IR) spectroscopy coupled with binary discrimination models was used for rapid and non-destructive discrimination. Among these models, the support vector machine (SVM) achieved 100% discrimination accuracy. For targeted authentication, data-driven soft independent modeling of class analogy (DD-SIMCA) one-class classification models were established under a compliant approach, and all four target classes achieved 100% sensitivity in the independent external test set. In addition, GC-MS analysis of volatile components revealed characteristic chemical features of the different medicinal parts and highlighted 20 compounds strongly associated with the root head (ASH). Notably, the root tail (AST) exhibited significantly higher antithrombin activity and a greater ability to prolong APTT than other parts. This integrated binary discrimination and one-class classification strategy offers a powerful tool for authenticating herbs and enhancing quality control in functional foods. • FT-IR spectroscopy and chemometrics effectively distinguish medicinal parts of Angelica Sinensis . • SVM model achieve 100% discrimination accuracy. • DD-SIMCA model shows high sensitivity and specificity in specific part certification. • The tail of Angelica Sinensis had significantly higher antithrombin activity.