Identifying active compounds and revealing integrated mechanism of phytomedicines via AI-driven chemical-biological information fusion: a case study of Weifuchun.

ETHNOPHARMACOLOGICAL RELEVANCE: Chinese phytomedicines exert holistic effects, yet their multidimensional mechanisms of action (MOAs) remain challenging to reveal. Standalone mass spectrometry (MS)-based chemical profiling is largely constrained to known targets, whereas transcriptomics-driven biological profiling relies on exhaustive searches and lacks compound-level specificity. Integrating chemical and biological information offers a systematic and generalizable approach to elucidating the coordinated MOAs underlying Chinese phytomedicines, such as Weifuchun, intended for chronic atrophic gastritis. AIM OF THE STUDY: To establish an AI-driven chemical-biological information fusion framework that systematically identifies active compounds and integrated MOAs of phytopharmaceuticals, using Weifuchun for demonstration. MATERIALS AND METHODS: MS data were combined with disease-related and transcriptomic data to construct chemical-biological information fusion networks. GCN-GAT-A∗, was developed to identify key action paths and novel targets. Molecular docking and in vitro/in vivo validations were performed to investigate compound-target interactions, signaling pathways, and pharmacological effects. RESULTS: WFC reduced gastric tissue IL-1β and IL-6 levels by 66.3% and TNF-α by 50.9% in the CAG model, and that in vitro treatment significantly decreased NO, IL-6, and TNF-α levels in LPS-stimulated macrophages. Moreover, naringenin, ginsenoside Rc and diosmetin were identified as core active compounds. They collectively regulated potential key targets (DUSP1, IRS1 and FLT1) and signaling pathways (PI3K/AKT, MAPK), leading to inflammation suppression, oxidative stress inhibition, intestinal metaplasia reversal, and mucosal repair. Experimental validations confirmed synergistic anti-inflammatory and antioxidant effects, and activation of gastric mucosal repair. Weifuchun exerts integrated multi-compound, multi-target, multi-pathway, and multi-channel effects through synergistic regulation. CONCLUSION: This study presents an intelligent virtual screening approach that leverages chemical-biological information fusion to enable efficient and accurate identification of novel active compounds and their targets, as well as elucidation of MOAs, thereby providing a robust methodological tool for phytomedicine research and botanical drug discovery.