Machine learning-assisted network pharmacology reveals that the Chaihu-Longgu-Muli decoction modulates the inflammatory microenvironment to treat perimenopausal syndrome

Abstract

Background: Chaihu-Longgu-Muli decoction (CLMD) is a traditional Chinese medicine formula that shows promise in alleviating symptoms related to premenstrual syndrome (PMS). However, the underlying mechanism remains unclear. This study uses a machine learning-assisted framework integrated with network pharmacology and experimental validation to elucidate the key targets and signaling pathways involved in the therapeutic effects of CLMD on PMS. Methods: We developed an integrative research framework that incorporates network pharmacology, machine learning, molecular dynamics, and in vitro validation. First, we built an overlap network by intersecting disease-related gene sets with data from the TCMSP, BATMAN-TCM, and other relevant databases. We subsequently performed GO and KEGG enrichment analyses. Second, we generated a protein‒protein interaction (PPI) network and screened key targets via machine learning algorithms. Third, we evaluated key active components and targets for ligand‒receptor binding stability via molecular docking and 200 ns MD simulations. Finally, we validated the proposed mechanism by assessing the ability of CLMD to modulate the inflammatory microenvironment using Raw264.7 cells as the experimental model. Results: By constructing an intersecting network of CLMD-active ingredient-disease targets, we identified 204 representative active components and nearly 300 potential targets. Intersecting these genes with PMS-related genes yielded 46 key targets. The PPI network built in Cytoscape/STRING, together with multiple machine learning algorithms (LASSO, SVM-RFE, and random forest), was used to select key targets, including IL6, TNF, and IL1B. At the molecular level, the key active components (quercetin, kaempferol, and wogonin) showed strong docking affinities to these targets (binding energies