Integrating non-targeted metabolomics and machine learning for comprehensive phytochemical profiling and intelligent discrimination of Acorus tatarinowii and its adulterants.

The dried rhizome of Acorus tatarinowii Schott (SCP) is a valued medicinal material in traditional Chinese medicine (TCM). However, its quality and efficacy are often compromised by adulteration with morphologically and chemically similar species, including Acorus calamus L. (ZCP), Acorus calamus f. submersa Glück (WCP), and Anemone altaica Fisch. (JJCP). To address this issue, we developed a robust authentication strategy by integrating ultra performance liquid chromatography quadrupole time of flight-tandem mass spectrometry (UPLC-Q-TOF-MS/MS)-based non-targeted metabolomics with machine learning (ML). Through comparison with authentic standards and high-confidence database matching, a comprehensive phytochemical profiling identified a total of 181 compounds across all studied species, with 141 characteristic compounds detected in SCP. Orthogonal partial least squares discriminant analysis (OPLS-DA) identified 40 key differential metabolites as potential biomarkers for species discrimination. Subsequently, six ML classification models were systematically constructed and evaluated. The support vector machine (SVM) model demonstrated optimal performance, achieving 100 % accuracy on both training and test sets in the internal validation, along with significantly improved computational efficiency compared to traditional chromatographic methods. Collectively, the integrated "UPLC-Q-TOF-MS/MS metabolomics-ML" workflow was successfully validated for authenticating SCP and its adulterants. This strategy proves to be both efficient and reliable, providing not only novel data support for SCP quality control but also offering a promising and transferable technical pathway for the accurate discrimination of closely related medicinal materials.