Metabolomics coupled with machine learning highlights metabolite profile differences of Scutellaria baicalensis Georgi from different specifications and geographical origins.

Scutellaria baicalensis Georgi (SBG) is recognized as a species with both medicinal and nutritional value, exhibiting extensive pharmacological activities and edible value. SBG from multiple origins and in two distinct specifications is widely available in the market at the present time. A rapid and accurate analytical method was developed by combining machine learning with metabolomics strategies to discriminate between two specifications and four sources of SBG, and to identify their respective differential markers (DMs). The establishment of a qualitative identification model to distinguish the geographical origins and specifications of SBG was established by employing cost-sensitive models, feature selection techniques, and machine learning algorithms. Both discriminant models demonstrated satisfactory performance. Nine DMs, including organic acids, isoflavones, fungal metabolites, and sesquiterpenes, were considered as potential chemical markers for discriminating the geographical origin of SBG. In contrast to the former, two SBG specifications (Kuqin and Ziqin), abbreviated to KQ and ZQ, contain a total of 15 DMs, which predominantly belong to the flavonoid components. The strategy proposed in this study provides foundation for constructing qualitative identification models, offers valuable insights into how geographical origins and morphological specifications shape the metabolic profiles of SBG and other food-medicine homologous herbs.