An Integrative Workflow Combining Machine Learning and GNPS to Identify Key Anti‐Inflammatory Constituents in Honey‐Processed Ephedrae Herba ( Ephedra sinica Stapf)

Ephedrae Herba (EH), a traditional Chinese medicine long used for asthma treatment, is commonly processed with honey to yield honey-processed Ephedrae Herba (HEH), a form believed to enhance therapeutic efficacy. This study developed an integrated approach combining bioactivity evaluation, mass spectrometry, and machine learning to identify the key anti-inflammatory constituents of HEH. HEH extract was separated into five fractions using column chromatography. Anti-inflammatory activity was assessed in lipopolysaccharides (LPS)-induced RAW 264.7 macrophages, with Fr C demonstrating the strongest inhibition of nitric oxide (NO), tumor necrosis factor-α (TNF-α), interleukin-1β (IL-1β), and IL-6 production. UPLC-quadrupole time-of-flight mass spectrometer (Q-TOF-MS) and Global Natural Products Social Molecular Networking (GNPS)-based molecular networking enabled the tentative identification of 150 compounds in Fr C. Using the machine learning model (InflamNat) and molecular docking, 14 compounds were predicted to show anti-inflammatory potential. Four high-priority compounds were selected based on docking scores and experimentally validated. Each significantly suppressed the release of NO and pro-inflammatory cytokines (p < 0.01). These results indicate that the enhanced anti-inflammatory activity of HEH is attributable to its flavonoid and alkaloid components, providing a chemical basis for its traditional use in inflammatory diseases such as asthma. Our findings not only clarify the material basis of HEH's but also demonstrate a powerful analytical strategy for rapidly identifying bioactive compounds in processed herbal medicines.