{"paper_id":"284538ac-4f4e-4828-9d5a-16252fb7d06f","body_text":"ARTICLE | Article in Press\nScientific Reports https://doi.org/10.1038/s41598-026-77411-x\nScientific Reports\nThis Article in Press is shared early to give you faster access to new research. It is citable and carries a permanent DOI. \nThe final edited version will replace it automatically.\nIntegrating bioinformatics analysis with molecular \nsimulation to elucidate the therapeutic mechanism of \nGuizhi Fuling Wan for adenomyosis and endometriosis\nJingyi Wu · Xinting Li · Shuikun Li · Yanting Luo · Ying Gao · Yanfei Tao · Jiangying Yang · \nWeiyong Hong\nReceived: 14 April 2026 / Accepted: 20 October 2026\n© The Author(s) 2026\nAbstract\nTo elucidate the mechanism of Guizhi Fuling Wan (GFW) against adenomyosis (AM) and endometriosis (EM) \nusing an integrated approach combining network pharmacology, molecular docking, and molecular dynamics \nsimulation. GFW components and their targets were retrieved from the Traditional Chinese Medicine Systems \nPharmacology Database and Analysis Platform (TCMSP) and SwissTargetPrediction databases, while AM and \nEM targets were obtained from GeneCards and Open Targets Platform. Intersection targets were identified, \nand a herb-ingredient-target network was constructed. Protein–protein interaction networks, Gene Ontology \n(GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses, and molecular docking \nwere performed, followed by 200 ns molecular dynamics simulations. Seventy-six active ingredients and 480 \npotential GFW targets were identified, with 176 intersecting with AM and EM. Core ingredients included \ncerevisterol, kaempferol, and baicalein; hub targets comprised AKT1, MAPK3, HSP90AA1, BCL2, TP53, \nESR1, and JUN. Enrichment analysis implicated PI3K-Akt, MAPK, focal adhesion, and apoptosis pathways. \nMolecular docking and dynamics confirmed stable interactions between core ingredients and AKT1. GFW \nmay exert therapeutic effects against AM and EM through a synergistic multi-component, multi-target mecha -\nnism involving the neuro-endocrine-immune-metabolic axis and key signaling pathways, which may provide a \nsystems-level rationale for its clinical application.\nKeywords Guizhi Fuling Wan · Adenomyosis · Endometriosis · Network pharmacology · GEO database · \nMolecular dynamics simulation\nIntroduction\nAdenomyosis (AM) and endometriosis (EM) are complex, estrogen-dependent gynecological disorders that share \nsignificant clinical and pathological features while exhibiting distinct differences 1. Both AM and EM can lead \nto severe symptoms such as dysmenorrhea, infertility, and menorrhagia, significantly impairing patients’ quality \nof life and reproductive health 2. Due to the similarity in pathological cell types with only differences in lesion \n\nArticle in Press\nScientific Reports\nhttps://doi.org/10.1038/s41598-026-77411-x\nlocation, they are referred to as “sister diseases” of the endometrium 3. AM is characterized by the invasion of \nendometrial glands and stroma into the myometrium, resulting in diffuse or focal thickening of the uterine wall \nand uterine enlargement4,5. Its associated dysmenorrhea typically coincides with the menstrual cycle and tends \nto worsen progressively. In contrast, EM is defined by the presence of endometrial-like tissue outside the uterine \ncavity, commonly involving the ovaries, pelvic peritoneum, and other extrauterine sites4,5. Based on lesion loca-\ntion and morphology, EM is categorized into three main subtypes: superficial peritoneal lesions (SUP), ovarian \nendometriomas (OMA), and deep infiltrating endometriosis (DIE)6. Symptoms vary with lesion site; for example, \nOMA is frequently associated with chronic pelvic pain and infertility 4, while lesions near the rectum or bladder \nmay cause dyschezia or dysuria7.\nAlthough AM and EM differ in anatomical location, emerging evidence suggests they may originate from \na common pathogenic pathway, representing different manifestations within the same disease spectrum 6,8–10. \nBoth disorders involve abnormal migration and implantation of endometrial cells11 and exhibit overlapping gene \nexpression profiles in endometrial tissue12. Clinically, AM and EM frequently coexist13–15; notably, DIE is often \ndetected in patients with AM 15, further supporting a shared etiological link. Single-cell transcriptomic studies \nhave also identified several key regulatory pathways related to hormone signaling and cellular homeostasis that \nare commonly dysregulated in both conditions16. Therefore, elucidating the similarities and distinctions between \nAM and EM is essential for advancing toward precise diagnosis and tailored therapeutic strategies.\nClinical therapeutic agents for AM and EM include oral contraceptives, high-efficiency progestogens, and \ngonadotropin-releasing hormone (GnRH) analogues. While these agents can alleviate symptoms and slow dis -\nease progression by modulating hormonal activity, they are often accompanied by adverse effects such as gas -\ntrointestinal discomfort and weight gain, and symptoms frequently recur after treatment cessation 11. In contrast, \nTraditional Chinese Medicine (TCM) adopts a holistic, multi-target approach, demonstrating clinically meaning-\nful efficacy with a favorable safety profile and minimal side effects, thereby offering a valuable complementary \noption in practice.\nThe pathological basis of AM and EM was classified into categories such as “Zheng Jia” (abdominal masses), \ninfertility, dysmenorrhea, and menstrual irregularities in TCM. Although syndrome differentiation may vary, \nblood stasis obstructing the Chong and Ren meridians and the uterus is considered the core pathogenesis. Conse-\nquently, promoting blood circulation and resolving stasis constitute the fundamental therapeutic principle.\nGuizhi Fuling Wan (GFW), a classical formula first documented in the Jin Gui Yao Lue  (Essentials of the \nGolden Cabinet), is renowned as the “foremost formula for dissipating masses.” It is widely used to treat various \ngynecological disorders, including AM, EM, uterine fibroids, and polycystic ovary syndrome. However, most \nexisting studies focus on clinical observations, while the pharmacodynamic material basis and core mechanisms \nof GFW remain poorly understood, limiting its modern application and evidence-based integration.\nTCM compounds are grounded in a holistic philosophical framework. Network pharmacology, with its sys -\ntems-level perspective and focus on multi-component interactions, methodologically aligns with this TCM para-\ndigm and offers a robust approach to investigating the mechanistic basis of complex formulas. In this study, we \npresent an innovative integration of multi-omics analysis, network pharmacology, and molecular dynamics simu-\nlations to investigate the synergistic mechanisms of GFW in treating AM and EM, with a particular focus on their \nshared pathological features. Addressing the critical clinical challenges of high recurrence rates associated with \nconventional therapies and the poorly characterized mechanisms of TCM interventions, we systematically elu -\ncidate the “multi-component, multi-target, multi-pathway, multi-system” action model of GFW. This integrated \napproach overcomes the inherent limitations of fragmented, reductionist investigations into TCM mechanisms \nby identifying the key pharmacodynamic constituents and their core molecular targets, thereby establishing a \nfoundation for evidence-based TCM formulation development. Furthermore, by delineating the core molecular \nmechanisms underlying GFW therapeutic effects, our findings provide a theoretical framework for optimizing \npersonalized treatment strategies, potentially reducing disease recurrence and ultimately improving clinical out-\ncomes for patients with AM and EM.\n\n\nArticle in Press\nScientific Reports\nhttps://doi.org/10.1038/s41598-026-77411-x\nMaterials and methods\nScreening of active ingredients and putative targets of GFW\nThe chemical constituents of Guizhi (GZ), Fuling (FL), Mudanpi (MD), Chishao (CS), and Taoren (TR) were \nobtained from the Traditional Chinese Medicine Systems Pharmacology Database and Analysis Platform \n(TCMSP). Candidate compounds were selected based on the following criteria: oral bioavailability (OB) ≥ 30% \nand drug-likeness (DL) ≥ 0.18. The SMILES notations of these compounds were retrieved from the PubChem \ndatabase and subsequently submitted to the SwissADME web server (http://www.swissadme.ch/) for  s e c o n d a r y - r \no u n d filtering. Compounds were retained if they exhibited high gastrointestinal (GI) absorption and satisfied at \nleast two of five drug-likeness rules (Lipinski, Ghose, Veber, Egan, and Muegge)17–20. The SMILES strings of the \nresulting active compounds were then input into the SwissTargetPrediction database  (   h t t p : / / w w w . s w i s s t a r g e t p r e \nd i c t i o n . c h /     ) to predict potential protein targets, with the species restricted to Homo sapiens. Targets with a prob-\nability score greater than zero were considered significant. Finally, all identified protein targets were standardized \nto official gene symbols using the UniProt database.\nCollection of disease targets\nGene expression datasets related to AM and EM were retrieved from the Gene Expression Omnibus (GEO) data-\nbase using the keywords “adenomyosis” and “endometriosis” with “Organism” set as Homo sapiens. Datasets \ncomprising at least 10 samples (minimum 5 cases and 5 controls) were selected for further analysis. The SV A \npackage in R was applied to correct batch effects across multiple datasets to eliminate expression bias derived \nfrom different platforms and batches. Differential analysis was performed according to data types: the limma \npackage was used for microarray data, while the limma-voom pipeline was adopted for RNA-seq data. Differ -\nentially expressed genes between control and disease samples were screened with unified thresholds of P < 0.05 \nand |log2FC| ≥ 0.5 (FC: Fold change).Target genes associated with AM and EM were retrieved from GeneCards \nand the Open Targets Platform using “adenomyosis” and “endometriosis” as search keywords. Differentially \nexpressed genes (DEGs) from multiple datasets for the same disease were first merged and deduplicated, and then \ncombined with targets obtained from public databases by taking the union set. Duplicate genes were removed \nwithout applying additional score cut-offs to construct complete disease target gene sets for AM and EM. The \nactive targets of GFW, AM-related targets and EM-related targets were imported into the Bioinformatics Analysis \nPlatform. Venn diagram analysis was conducted to obtain the intersecting targets among GFW-AM-EM, and the \nVenn diagram was visualized.\nConstruction of the “Herb-Active Component–Target” network\nThe active components of GFW and their corresponding intersection targets were imported into Cytoscape (ver-\nsion 3.10.3) to construct an integrated “herb-active component-target” network, illustrating the potential mecha-\nnistic framework underlying GFW in treating AM and EM. Network topological parameters were calculated \nusing the built-in Network Analyzer tool, and active components with high degree values were identified as core \ncomponents for GFW treatment of AM and EM. Furthermore, toxicity prediction for these screened core com -\nponents was predicted using the ProTox 3.0 platform (https://tox.charite.de/) to evaluate potential safety profiles.\nConstruction of protein–protein interaction network and extraction of core targets\nThe intersection targets were submitted to the STRING database ( https://string-db.org) to construct a protein–\nprotein interaction (PPI) network. The minimum required interaction score was set to “highest confidence” (0.9), \nisolated nodes were hidden, and all other parameters were kept at default settings. The resulting network data in \n\n\nArticle in Press\nScientific Reports\nhttps://doi.org/10.1038/s41598-026-77411-x\nTSV format were imported into Cytoscape (version 3.10.3). Comprehensive topological analysis was conducted \nusing four algorithms (Degree, EPC, MCC, and MNC) implemented in the CytoHubba plugin. Targets commonly \nidentified by all four algorithms were determined via Venn diagram analysis and defined as core targets.\nTissue distribution characteristics of core targets\nThe expression of core targets across various normal human tissues was retrieved from the BioGPS database \n(http://biogps.org/). A “core target-tissue” distribution network was subsequently constructed using Cytoscape \n(version 3.9.1) to visualize the tissue enrichment characteristics of these targets under physiological conditions.\nFunctional association network analysis of core targets\nFunctional association network analysis of core targets was performed using a multi-database integration strat -\negy. First, a functional association network incorporating multi-dimensional data-including protein physical \ninteractions, co-expression, and pathway co-enrichment-was constructed using the GeneMANIA database  (   h t t p s : \n/ / g e n e m a n i a . o r g /     ) . This analysis revealed functional synergistic relationships and potential functional modules of \ncore targets at the systems level. Simultaneously, the IntAct database (https://www.ebi.ac.uk/intact/) and HIPPIE \ndatabase (https://cbdm-01.zdv.uni-mainz.de/~mschaefer/hippie/) were utilized for complementary validation and \nquantitative assessment of PPI involving core targets, thereby enhancing the reliability and comprehensiveness \nof the findings.\nConstruction of multi-level regulatory networks of core targets\nTo further explore the regulatory mechanisms of core targets, the TRRUST database was utilized to analyze \ninteractions between transcription factors (TFs) and core targets and to predict cascade regulatory relationships \namong TFs. Subsequently, these TF-gene and miRNA-gene interaction datasets were integrated to construct a \nmulti-level “TF-miRNA-mRNA” regulatory network using Cytoscape (version 3.10.3), systematically revealing \nthe hierarchical regulatory landscape of core targets in AM and EM.\nGO and KEGG enrichment analysis of intersection targets\nGene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were \nperformed using the DA VID (Database for Annotation, Visualization and Integrated Discovery) database to \ninvestigate the biological processes (BP), cellular components (CC), molecular functions (MF), and signaling \npathways associated with the intersection targets of GFW against AM and EM. “OFFICIAL GENE SYMBOL” \nwas selected as the identifier, “Gene List” as the list type, and Homo sapiens as both the species and background. \nA significance threshold of P < 0.05 was applied. The GO and KEGG enrichment results were visualized using a \nbioinformatics analysis platform.\nMolecular docking validation\nTo validate the reliability of our findings and to elucidate the interaction patterns and binding affinities between \ncore active components and core targets, semi-flexible molecular docking was performed. The detailed workflow \nwas as follows: First, acquisition and structural optimization of small-molecule ligands were performed. The \n.mol2 files of core active components were retrieved and downloaded from the TCMSP. These files were con -\nverted into Protein Data Bank (PDB) format via PyMOL software. Geometric optimization was carried out using \nAvogadro software with the MMFF94s force field to correct unreasonable structures such as bond lengths and \nbond angles, so as to guarantee the structural rationality of small-molecule ligands. Second, three-dimensional \n\n\nArticle in Press\nScientific Reports\nhttps://doi.org/10.1038/s41598-026-77411-x\nstructures of core target proteins were obtained and preprocessed. Three-dimensional crystal structures of core \ntarget proteins including AKT1, MAPK3, HSP90AA1, TP53, JUN, BCL2, ESR1 were retrieved and downloaded \nfrom the PDB21–27. Standardized preprocessing was conducted for all protein structures, mainly including removal \nof water molecules and irrelevant heteroatoms, hydrogen-atom addition, and charge assignment. Subsequently, \nthe protein active binding regions and docking box parameters were determined. Site-specific docking mode was \nadopted. For protein structures with co-crystallized ligands, the active binding region was defined according to \nthe position of ligands in the original crystal structure. For structures without co-crystallized ligands, the Pro -\nteinsPlus web server was used to predict potential functional binding pockets, and the molecular docking region \nwas determined based on the prediction results. Meanwhile, the GetBox plugin in PyMOL was utilized to obtain \ndocking box parameters for each target protein, including central coordinates (center_x, center_y, center_z) and \nbox dimensions (size_x, size_y, size_z). Detailed parameters are shown in Table S1. Finally, docking parameters \nwere set, followed by conformation screening and interaction analysis. Uniform molecular docking parameters \nwere set as follows: exhaustiveness for conformational search was set to 8; 20 candidate binding poses were \ngenerated for each ligand (num_modes = 20); the maximum energy difference among output conformations was \nset to 4 kcal/mol (energy_range = 4). After docking, optimal binding conformations were selected according to \nbinding free-energy values, spatial fit of ligands within protein binding pockets, and interaction profiles between \nligands and key amino acid residues of proteins. Discovery Studio software was applied for two-dimensional \ninteraction analysis of the optimal binding conformations. Non-covalent interactions between ligands and target \nproteins were systematically explored, including hydrogen bonds, van der Waals forces, π–π stacking, π–alkyl \ninteractions, π–anion interactions, and π–σ interactions. Key amino acid residues involved in interactions and \nhydrogen-bond distances were accurately recorded to clarify the binding mechanism between components and \ntargets.\nAfter all ligand-receptor docking calculations were completed, representative targets for molecular dynamics \nsimulations were screened according to three pre-established criteria. (i) quantitative binding criterion: the target \ncan form stable complexes with all six core active components, with ligand-receptor binding free energy ≤ − 7 \nkcal/mol28; (ii) comparative analysis criterion: parallel comparison of all active components can be realized to \nintuitively explore the binding interactions between multiple components from traditional Chinese medicine and \na single protein target ; (iii) pathway-level criterion: priority is given to upstream core regulatory proteins cor -\nresponding to the top three pathways in terms of enrichment significance from KEGG enrichment analysis. After \nscreening, one representative receptor was determined, and the optimal docking conformation of each complex \nwas extracted for subsequent molecular dynamics simulations.\nMolecular dynamics simulation\nBuilding upon the molecular docking results, 200 ns molecular dynamics (MD) simulations were performed \nusing the GROMACS (version 2021.6) software package to systematically investigate the binding systems of six \nsmall molecules with the AKT1 protein and to further evaluate the dynamic stability and interaction mechanisms \nof the ligand-receptor complexes. The AKT1 protein was parameterized using the Amber14sb force field. Small-\nmolecule ligands underwent structural optimization using the r2SCAN-3c method implemented in the ORCA \nprogram, followed by single-point energy calculations at the B3LYP-GD3/DEF2-TZVP level of theory combined \nwith the def2/J auxiliary basis set. Atomic charge were fitted using the RESP2 method in Multiwfn, and topology \nfiles were generated based on the GAFF2 force field. All simulation systems were solvated in a cubic water box \nusing the TIP3P water model, ensuring appropriate distance between the solute and box boundaries. Counterions \nwere added to neutralize the system charge. Energy minimization was first performed using the steepest descent \nalgorithm until the maximum force was below 1000 kJ/mol/nm, followed by conjugate gradientminimization to \na maximum force of 100 kJ/mol/nm. Sequential equilibration steps were subsequently conducted in the canoni -\ncal (NVT) ensemble at 310.15 K for 1 ns, followed by the isothermal-isobaric (NPT) ensemble at 310.15 K and \n1.0 bar for 1 ns to estabilize temperature and pressure. Production MD simulations were carried out for 200 ns. \n\n\nArticle in Press\nScientific Reports\nhttps://doi.org/10.1038/s41598-026-77411-x\nAll bonds containing hydrogen atoms were constrained using the LINCS algorithm, long-range electrostatic \ninteractions were treated using the Particle Mesh Ewald (PME) method, a time step of 2 fs was employed, and \ntrajectory snapshots recorded every 1.0 ps. The MM/GBSA method was adopted to calculate the binding free \nenergies between candidate compounds and target proteins. Binding free-energy calculations were performed \nusing equilibrated conformations derived from 200 ns molecular dynamics simulation trajectories. Considering \nthat obvious structural rearrangements still occurred in the early simulation phase, the stable trajectories from the \nfinal 20 ns (180–200 ns) were selected for subsequent analysis. Representative conformations were extracted at \na time interval of 100 ps, yielding a total of 200 snapshots for MM/GBSA binding free-energy computation. The \nfinal binding free energy and per -residue energy decomposition results were obtained by averaging the calcu -\nlated values over all representative snapshots, so as to reduce the influence of conformational fluctuations from \nindividual snapshots on the computational outcomes. The ProLIF package was used to extract protein-ligand \ninteraction fingerprints across the entire simulation trajectory, so that the occurrence frequency and stability of \nkey intermolecular interactions could be systematically evaluated.\nResults\nScreening of active ingredients and targets of GFW\nThe chemical constituents of GZ, FL, MD, CS, and TR were retrieved from the TCMSP. The first-round screen-\ning was carried out with the thresholds of OB ≥ 30% and DL ≥ 0.18. The number of compounds obtained from the \nprimary screening of each herb was listed as follows: MD: 11, GZ: 7, FL: 15, TR: 23, CS: 29. The cumulative \nnumber of compounds from the primary screening of the five herbs reached 85, including overlapping constitu -\nents shared by multiple herbs. After removing duplicate compounds, 76 non-redundant chemical constituents \nunique to the formula were acquired. These 76 compounds were imported into the SwissADME platform for \nsecondary druggability screening. The screening criteria were set as high gastrointestinal absorption, and at least \n2 of the five drug-likeness evaluation rules were judged as Yes. After the two-step progressive druggability \nscreening, a total of 48 major active constituents were obtained. The SMILES of the 48 active constituents were \nsubsequently input into SwissTargetPrediction for target gene prediction, followed by standardization via the \nUniProt database. Fifteen constituents without predicted target genes were eliminated, and 33 active constituents \nwere finally retained. After deleting duplicate target entries, 480 corresponding targets were acquired.\nCollection of disease targets\nThe GEO database was searched using “adenomyosis” as the keyword. After screening according to sample type, \ndata completeness and rationality of experimental design, three datasets were finally included: GSE193928 (19 \ncontrol cases and 6 AM cases), GSE190580 (5 control cases and 6 AM cases), and GSE185392 (10 control cases \nand 10 AM cases).The GEO database was searched with “endometriosis” as the keyword, and four datasets were \nincluded following the same screening criteria: GSE212787 (6 control cases and 7 EM cases), GSE192636 (5 \ncontrol cases and 5 EM cases), GSE51981 (34 control cases and 114 EM cases), and GSE6364 (16 control cases \nand 21 EM cases). Raw transcriptomic data from all datasets were uniformly cleaned and filtered to construct \nstandardized expression matrices. The SV A package was applied to correct cross-dataset batch effects. Differen-\ntial expression analysis was performed using the limma package for array data and the limma-voom pipeline for \nRNA-seq data, with unified screening thresholds of P < 0.05 and |log2FC| ≥ 0.5. DEGs from different datasets for \nthe same disease were merged and deduplicated, yielding 5058 DEGs for AM and 11,500 DEGs for EM.\nTo achieve comprehensive coverage of disease-related genes, AM-DEGs were merged with AM-associated \ntargets retrieved from GeneCards and the Open Targets Platform, followed by deduplication, and a total of 5790 \nAM disease targets were obtained. Similarly, EM-DEGs were integrated with EM-related targets from public \n\n\nArticle in Press\nScientific Reports\nhttps://doi.org/10.1038/s41598-026-77411-x\ndatabases to generate 14,463 total EM disease targets. Only duplicate genes were removed during integration, and \nno additional filtering thresholds were applied. Intersection analysis among the AM targets (5790), EM targets \n(14463), and GFW potential targets (480) identified 176 common targets, which were visualized using a Venn \ndiagram (Fig. 1A).\nConstruction of the “Herb-Active Ingredient-Target” network and screening of core components\nThe active ingredients and intersection targets were imported into Cytoscape (version 3.10.3) to construct a \nvisualized herb-active ingredient-target network (Fig. 1B). The network comprised 215 nodes (1 formula node, 5 \nherb nodes, 33 active ingredient nodes, and 176 target nodes) connected by 700 edges. Seven active ingredients \nwere found to interact with more than 30 targets. Notably, cerevisterol was linked to 46 targets, while popu -\nloside_qt and GA120 each interacted to 40 targets. Using the Network Analyzer tool, core active ingredients \nwere identified based on degree values, including cerevisterol, populoside_qt, kaempferol, baicalein, GA120, \nand (2R)-2-[(3 S,5R,10 S,13R,14R,16R,17R)-3,16-dihydroxy-4,4,10,13,14-pentamethyl-2,3,5,6,12,15,16,17-\noctahydro-1 H-cyclopenta[a]phenanthren-17-yl]-6-methylhept-5-enoic acid (hereafter referred to as enoic acid). \nToxicity prediction using the ProTox 3.0 platform confirmed that all core components exhibited LD₅₀ values \nexceeding 2000 mg/kg, indicating high safety profiles and classification as non-toxic (Table 1). These findings \nindicate that GFW exerts therapeutic effects against AM and EM through multiple components acting on multiple \ntargets.\nFig. 1 Target identification, network \nconstruction, and PPI analysis of \nGFW in AM and EM treatment. \n(A) Venn diagram of intersection \ntargets. (B) Herb-active ingredient-\ntarget interaction network. (C) PPI \nnetwork.\n \n\n\nArticle in Press\nScientific Reports\nhttps://doi.org/10.1038/s41598-026-77411-x\nScreening of core targets and PPI network analysis\nThe PPI network (Fig. 1C) was analyzed using four algorithms (Degree, EPC, MCC, and MNC) implemented in \nthe CytoHubba plugin of Cytoscape to screen the top 10 genes from each method. Common targets identified by \nall four algorithms were determined using Venn diagram analysis and designated as core targets: AKT1, MAPK3, \nHSP90AA1, TP53, JUN, BCL2, and ESR1 (Fig. 2).\nGeneMANIA network analysis revealed that the six core targets ( AKT1, MAPK3, JUN, HSP90AA1, BCL2, \nand ESR1) form a highly interconnected interaction network (Fig. 3A). The inner circle represents core targets, \nwhile the outer circle represents functionally related genes. Connections between nodes indicate PPI, with node \ncolors reflecting additional functional attributes. The network was dominated by physical PPI (70.21%), supple-\nmented by pathway associations (18.13%), co-expression relationships (9.37%), and co-localization interactions \n(1.86%), collectively underscoring the functional synergy among these genes. Although TP53 was not integrated \ninto this unified network due to database limitations, independent analyses using the IntAct and HIPPIE databases \nconfirmed its interaction with the other six core genes. This finding is consistent with prior literature documenting \nTable 1 Core active ingredients of GFW against AM and EM.\nMOL ID Molecule Name Degree OB/% DL LD50/mg/kg Source\nMOL000279 Cerevisterol 46 37.96 0.77 2340 FL\nMOL001371 Populoside_qt 40 108.89 0.2 4000 TR\nMOL001340 GA120 40 84.85 0.45 5000 TR\nMOL000273 (2R)-2-[(3 S,5R,10 S,13R,14R,16R,17R)-\n3,16-dihydroxy-4,4,10,13,14-\npentamethyl-2,3,5,6,12,15,16,17- \noctahydro-1 H-cyclopenta[a]\nphenanthren-17-yl]-6-methylhept-5-enoic \nacid\n38 30.93 0.81 4000 FL\nMOL002714 Baicalein 36 33.52 0.21 3919 CS\nMOL000422 Kaempferol 33 41.88 0.24 3919 MD\nActive components in this table were screened with thresholds of OB ≥ 30% and DL ≥ 0.18. Meanwhile, the top six core active molecules ranked \nby Degree in the PPI network were retained. MOL ID: molecular identifier from the TCMSP database; Degree: target connectivity of active \ncomponents in the protein-protein interaction network; OB/%: Oral Bioavailability; DL: Drug -likeness score; LD50/mg/kg: median lethal dose, \nrepresenting the acute toxicity of compounds;\nSource column abbreviations FL/TR/CS/MD: abbreviations of individual herbs in the GFW formula.\nFig. 2 Screening of core targets \nusing four algorithms.\n \n\n\nArticle in Press\nScientific Reports\nhttps://doi.org/10.1038/s41598-026-77411-x\ndirect interactions between TP53 and AKT1, MAPK3, JUN, HSP90AA1, BCL2, and ESR129–32, further supporting \nTP53 as a critical component of this functional network.\nConstruction of the “Core Target-Organ Tissue” network\nCore targets were imported into the BioGPS database with the species set as “human”. The top 10 organs and \ntissues with the highest gene expression were queried in the Interactive Image plot. A “core target-organ-tissue” \nnetwork was constructed using Cytoscape 3.9.1 (Fig. 3B). Network statistics and topological feature analysis \nrevealed that the seven core targets corresponded to 39 distinct human tissues and cell types. The connectivity of \neach target was relatively balanced, with each target associating with approximately 10-11 tissues. JUN was asso-\nciated with 11 tissues or cells, whereas TP53, AKT1, MAPK3, HSP90AA1, BCL2 and ESR1 each corresponded to \n10 tissues or cells. No tissue-target bias was observed, in which one single target was highly enriched while other \ntargets showed restricted distribution. Numerous shared targets existed across different physiological systems \nincluding immune, endocrine, circulatory, nervous and digestive systems, and no obvious modular separation \nwas detected in the network, indicating widespread overlapping expression of core regulatory molecules across \nhuman organs. This distribution feature only provides bioinformatic clues for the multi-organ regulatory potential \nof GFW based on basal gene expression profiles of healthy human tissues, and cannot directly confirm the phar-\nmacological effect of multi-organ synergistic regulation. Further statistical analysis showed that these 39 tissues \ncovered multiple human physiological systems, and the distribution characteristics of targets in each system were \nas follows. The blood and immune system contained the largest number of associated targets. Six out of seven \ncore targets (85.7%) were expressed in relevant cells and tissues of this system, including key immune cells such \nas CD33+_Myeloid, CD34+, CD56+_NKCells, CD19+_BCells, as well as whole-blood tissue. Four core targets \n(57.1%) were distributed in the endocrine system, mainly in endocrine organs such as the thyroid, hypothalamus \nand pineal gland, among which ESR1 served as a representative hormone-related target in this system. Six core \ntargets (85.7%) were found in the digestive system (liver, small_intestine, colon, etc.) and cardiovascular sys -\ntem (heart, cardiac myocytes, etc.). Five core targets (71.4%) were present in the nervous system (whole brain, \nprefrontal cortex, hypothalamus, etc.), collectively forming a pattern of widespread systemic expression of these \ntargets.\nTheoretically, EM and AM are categorized as gynecological blood-stasis syndrome in traditional Chinese \nmedicine. Their pathological mechanisms are not limited to local blood stasis in the uterus, and are often accom-\npanied by systemic homeostatic disturbances of the Chong-Ren meridian, immunity and endocrine function. The \nFig. 3 Holistic regulatory networks of GFW core targets across molecular, cellular, and systemic levels. ( A) GeneMANIA \nanalysis of core targets. ( B) Core target-organ tissue network. Hub genes were represented by orange nodes, and tissues/\norgans were represented by green triangles. ( C) TF-miRNA-mRNA regulatory network. Core targets were represented by \norange nodes, transcription factors were represented by green triangles, and miRNAs were represented by yellow polygons.\n \n\n\nArticle in Press\nScientific Reports\nhttps://doi.org/10.1038/s41598-026-77411-x\nwidespread cross-system overlapping distribution of core targets in our results can serve as indirect reference for \nunderstanding the holistic therapeutic strategy for gynecological blood-stasis syndrome in TCM. Nevertheless, \nthis inference remains a hypothesis derived purely from bioinformatic data and lacks supporting evidence from \nin vivo experiments such as multi-tissue protein detection and in-vivo pharmacological assays. Targeted experi-\nmental validation is required in further studies.\nConstruction of the TF-miRNA-mRNA regulatory network\nUsing the TRRUST and miRWalk databases, 18 upstream transcription factors and 8 targeting miRNAs were \npredicted based on the seven core targets ( TP53, AKT1, MAPK3, JUN, HSP90AA1, BCL2, and ESR1). A tran-\nscription factor-miRNA-mRNA regulatory network was constructed, comprising 10 miRNA-mRNA pairs and 37 \nTF-mRNA pairs (Fig. 3C). Network analysis revealed that TP53, JUN, and ESR1 serve not only as core targets \nbut also as key transcription factors with hub regulatory roles. Specifically, TP53 regulates targets such as BCL2 \nand AKT1; JUN acts on TP53 and BCL2; and ESR1 broadly regulates multiple genes including TP53, JUN, and \nBCL2. Through these direct transcription factor-core target regulatory relationships, these three factors govern \nthe balance of cell apoptosis, proliferation, and estrogen signaling pathways. Simultaneously, as core targets, \nthey are regulated by other transcription factors. TP53 is regulated by E2F1, HDAC1, and ESR1. JUN is targeted \nby ABL1, CREB1, and ESR1. ESR1 forms regulatory loops with PGR, HDAC1, and TWIST1. This dual role of \n“regulated-regulating” renders these three factors critical nodes connecting different signaling pathways, enhanc-\ning network synergy and signal integration capabilities. The remaining non-transcription factor targets ( AKT1, \nMAPK3, HSP90AA1, and BCL2) are also subject to direct regulation by TP53, JUN, and ESR1, participating in a \nglobal regulatory system encompassing cell proliferation, apoptosis, and inflammatory responses.\nBeyond the dominant roles of key transcription factors, multiple specific miRNA regulatory axes targeting \ncore targets were identified. Notably, hsa-miR-139-5p simultaneously targets both BCL2 and HSP90AA1, while \nhsa-miR-149-3p concurrently targets TP53 and AKT1. These findings suggest that GFW may achieve multi-target \nsynergistic intervention through individual miRNAs. Together with transcription factors, these miRNAs consti -\ntute a “transcriptional-post-transcriptional” dual-track regulatory model, enabling precise and dynamic modula-\ntion of core targets. The regulatory network clearly demonstrates that individual transcription factors (such as \nESR1, ABL1, and CREB1) can regulate multiple core targets, while individual core targets (such as TP53 and \nBCL2) receive signal inputs from multiple TFs and miRNAs. This “many-to-many” regulatory pattern suggests \nthat GFW does not act on isolated individual targets. Instead, by interfering with these highly interconnected \nnodes, it may modulate key pathways related to apoptosis, proliferation, inflammation and hormone response. \nGFW exhibits the potential to mediate the restoration of organismal homeostasis at the systemic level, which \nprovides preliminary molecular -network-level evidence supporting the TCM therapeutic principle of “holistic \nregulation”.\nGO and KEGG pathway enrichment analysis\nGO enrichment analysis of the intersection targets was performed (P < 0.05), yielding 598 BP terms, 73 CC terms, \nand 198 MF terms. BP terms were primarily enriched in positive regulation of phosphatidylinositol 3-kinase/pro-\ntein kinase B signal transduction, negative regulation of gene expression, positive regulation of cell population \nproliferation, negative regulation of apoptotic process, and inflammatory response. CC terms were predominantly \nassociated with cytosol, plasma membrane, protein-containing complex, nucleoplasm, and cytoplasm. MF terms \nwere mainly enriched in enzyme binding, nuclear receptor activity, endopeptidase activity, protein kinase activ -\nity, and protein serine kinase activity (Fig. 4A). KEGG pathway enrichment analysis identified 151 pathways. \nThe results indicated that GFW may exert therapeutic effects primarily through the PI3K-Akt signaling pathway, \nMAPK signaling pathway, endocrine resistance, focal adhesion, and apoptosis pathways. Key targets promi -\nnently involved in these pathways included AKT1 and MAPK3 (Fig. 4B).\n\n\nArticle in Press\nScientific Reports\nhttps://doi.org/10.1038/s41598-026-77411-x\nMolecular docking\nMolecular docking was performed between the core targets (AKT1, MAPK3, HSP90AA1, TP53, JUN, BCL2, and \nESR1) and the core active ingredients of GFW (cerevisterol, populoside_qt, GA120, enoic acid, baicalein, and \nkaempferol) to preliminarily validate binding activity (Fig. 4C). In molecular docking analysis, binding energy \nbelow 0 implies a preferential binding tendency between the ligand and receptor, with lower binding energies \ncorresponding to more stable conformations, while binding energy ≤ -5 kcal/mol indicates favorable binding \nactivity. Heatmap results demonstrated that binding free energies between all active ingredients and most targets \nwere below − 5 kcal/mol, indicating good binding potential. Certain ingredients displayed particularly strong \nbinding capabilities toward specific targets. For instance, cerevisterol with AKT1, kaempferol with MAPK3, and \npopuloside_qt with ESR1 all exhibited markedly low binding energies, suggesting efficient molecular interac -\ntions. Furthermore, baicalein and kaempferol exhibited strong binding activities toward multiple targets ( AKT1, \nHSP90AA1, and MAPK3), suggesting that their molecular structures could adapt to the binding pockets of mul -\ntiple targets, thereby possessing potential for multi-pathway regulation. Collectively, cerevisterol, baicalein, and \nkaempferol demonstrated the strongest binding activities and broadest target spectra among the candidate active \ningredients, particularly showing significant binding advantages toward key signaling pathway proteins including \nAKT1, MAPK3, and HSP90AA1.\nTo screen the optimal target for subsequent structural analysis and molecular dynamics simulation, systematic \nscreening of the seven core targets was performed based on the three-tier evaluation system established in Sect. \n2.9, as summarized in Table S2. The top three KEGG pathways with the highest enrichment significance ranked \nby p-value were PI3K-Akt signaling pathway, Endocrine resistance, and Focal adhesion. According to the com-\nprehensive screening results, both AKT1 and HSP90AA1 exhibited stable binding affinity with all six core active \ncompounds and satisfied the requirements for parallel comparative analysis. However, HSP90AA1 participates \nonly in the PI3K-Akt signaling pathway, whereas AKT1 acts as a shared upstream hub regulator of all three top-\nenriched pathways. Given its superior core regulatory status in the pathological network of AM and EM, AKT1 \nwas finally selected as the optimal target for subsequent binding pattern visualization and molecular dynamics \nsimulation.\nTo further explore the binding patterns between active components and the target protein, the docking results \nof AKT1 with six active components were selected for three-dimensional conformational visualization using \nPyMOL (Fig. 5). Furthermore, two-dimensional interaction analysis of the optimal docking conformations was \nperformed via Discovery Studio to systematically characterize the interaction modes between ligands and key \nFig. 4 Enrichment analysis and molecular docking validation of GFW therapeutic targets. (A) Gene Ontology (GO) enrich-\nment analysis. (B) Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis, performed based on \nthe KEGG database (Kanehisa Laboratories) under CC BY 4.0 license with formal permission; see reference33. (C) Heatmap \nof molecular docking binding energies.\n \n\n\nArticle in Press\nScientific Reports\nhttps://doi.org/10.1038/s41598-026-77411-x\namino acid residues within the protein active pocket. The detailed interaction parameters between each com -\npound and AKT1 are summarized in Table S3. The results indicated that all six active components could sta -\nbly embed into the active cavity of AKT1 and bind to key residues through multiple non-covalent interactions, \nincluding hydrogen bonds, van der Waals forces, hydrophobic interactions, and π-related interactions, thereby \nFig. 5 Visualization of molecular \ndocking results. (A) AKT1 with \nbaicalein. (B) AKT1 with GA120. \n(C) AKT1 with enoic acid. (D) \nAKT1 with Cerevisterol. (E) AKT1 \nwith populoside_qt. (F) AKT1 \nwith kaempferol. (G) AKT1 with \nCapivasertib.\n \n\n\nArticle in Press\nScientific Reports\nhttps://doi.org/10.1038/s41598-026-77411-x\nmaintaining the stability of ligand-receptor complexes. Specifically, baicalein bound to AKT1 with a binding \nenergy of − 9.7 kcal/mol. It formed one hydrogen bond with THR211 at a bond length of 3.1 Å. In addition, baica-\nlein generated van der Waals forces with SER205, ALA212, ILE290, TYR272, THR291 and ASN53, and formed \na π-anion interaction with ASP292. Residues including TRP80, LEU264, V AL270, LYS268 and LEU210 partici-\npated in π-π stacking, π-σ and hydrophobic interactions. These multiple intermolecular forces jointly contribute \nto the stable binding of the ligand-receptor complex. Cerevisterol exhibited one of the strongest binding affinities \ntoward AKT1, with a binding energy of − 11.3 kcal/mol. It formed a hydrogen bond with TYR272 at a bond length \nof 2.8 Å. Extensive van der Waals interactions were observed with ASN54, V AL271, ASP274, GLN79, ARG273, \nILE84, SER205, TYR263, CYS296, THR82 and ASP292. Residues TYR272, LEU264, TRP80, LYS268 and \nV AL270 mediated hydrophobic interactions, suggesting that this molecule possesses favorable spatial compat -\nibility with the protein binding pocket.GA120 had a binding energy of − 8.9 kcal/mol against AKT1. It formed \nfour hydrogen bonds with THR82 (3.1 Å), THR81 (2.8 Å), V AL271 (2.7 Å) and ASN54 (3.1 Å), respectively. \nVan der Waals interactions were detected with ASP292, TRP80, GLY294, GLN79, ASP274, CYS296, ILE84 and \nARG273, while TYR272 and V AL270 were involved in hydrophobic interactions. Kaempferol bound to AKT1 \nwith a binding energy of − 9.1 kcal/mol through four hydrogen bonds, which were formed with THR211 (2.9 Å), \nASN54 (3.1 Å), V AL271 (3.3 Å) and ILE290 (3.2 Å). It exerted van der Waals forces with SER205, TYR263, \nLYS268, GLN79, ARG273 and TYR270. Residues including ASP292, V AL270, TRP80, LEU210 and LEU264 \nmediated π-anion, π-σ, π-π stacking and hydrophobic interactions, indicating the formation of a stable binding \nsystem between kaempferol and AKT1. Populoside_qt showed a binding energy of − 8.3 kcal/mol with AKT1 and \nformed one hydrogen bond with ILE290 at a bond length of 3.2 Å. Multiple van der Waals and hydrophobic inter-\nactions were observed with GLN79, V AL271, V AL270, TRP80, ASP292, THR211, THR291, TYR272, THR82, \nILE84 and ASN54. Enoic acid exhibited a binding energy of − 8.1 kcal/mol with AKT1. It formed three hydro -\ngen bonds, including two bonds with ASP325 (3.3 Å and 2.9 Å) and one bond with LYS39 (3.1 Å). Residues \nMET363, LYS389, GLU319, ASP387, TYR38, ARG41, GLU40, ASN324, LEU362 and GLU322 provided van \nder Waals interactions, whereas LYS386, PRO388 and PRO42 participated in hydrophobic interactions. In sum-\nmary, all tested active components can stably anchor within the AKT1 active pocket through diverse non-cova -\nlent interactions. Hydrogen bonds and hydrophobic interactions are presumed to be the core driving forces for \nmaintaining complex stability. Among these components, cerevisterol, baicalein and kaempferol presented lower \nbinding energy values and established abundant molecular interaction networks with key amino acid residues of \nAKT1, indicating their superior binding potential with the target protein.\nTo objectively evaluate the target-binding capacity of the six candidate components, the clinically available \nAKT1 inhibitor capivasertib was employed for molecular docking to compare the interaction modes and bind -\ning properties of the candidate components and the reference drug with the AKT1 receptor. Capivasertib is an \noral AKT kinase inhibitor 19,20. The docking simulation results of capivasertib with AKT1 were integrated into \nFig. 5 for synchronous comparison with the docking conformations of the six candidate components. The dock -\ning results revealed that its binding mode includes hydrogen bonds with TYR324 and TRP78, accompanied by \nTRP78-mediated π-π stacking interactions. Residues including THR79, V AL269, THR195, V AL199, GLY325, \nASP323 and ASN267 stabilize the complex structure via van der Waals forces. Visualization showed that capiv-\nasertib binds to the hydrophobic pocket of the AKT1 protein; the six candidate components also bind within this \nhydrophobic pocket, with highly overlapping binding regions relative to capivasertib, verifying the reliability of \nthis docking system and the comparability of results across groups. Notably, the binding free energies of the six \ncandidate small molecules to AKT1 were all lower than that of capivasertib. In the molecular docking evaluation \nframework, a lower binding free energy indicates stronger affinity between the ligand and target and higher stabil-\nity of the complex, suggesting that the six active ingredients screened from GFW exhibit superior AKT1-binding \ncapacity compared with the reference inhibitor and possess promising target-binding potential.\n  \n\n\nArticle in Press\nScientific Reports\nhttps://doi.org/10.1038/s41598-026-77411-x\nMolecular dynamics simulation\nBased on molecular docking results, 200 ns molecular dynamics simulations were performed for each complex \nto evaluate binding stability and dynamic interactions (Fig. 6A and D). Root-mean-square deviation (RMSD) \nserves as a core indicator for assessing conformational stability during MD simulations, with values reflecting \nFig. 6 Molecular dynamics simulations and binding free energy analysis of GFW active ingredients with AKT1. (A) RMSD \nof AKT1 protein in different complex systems. (B) RMSD of small molecule-AKT1 complexes. (C) RMSF of AKT1 protein \nresidues in different complex systems. (D) Hydrogen bond numbers between small molecules and AKT1 protein. (E) Bind-\ning free energies and energy decomposition of six candidate small molecules with AKT1 protein calculated by MM/GBSA.\n \n\n\nArticle in Press\nScientific Reports\nhttps://doi.org/10.1038/s41598-026-77411-x\nthe deviation between the dynamic conformation and initial structure. Smaller RMSD values indicate weaker \nconformational fluctuations and higher stability. Figure 6A and B illustrate the RMSD trends of AKT1 protein \nand the overall complexes across different systems, respectively. The RMSD of AKT1 protein directly reflects \nthe impact of various small molecule bindings on protein conformational fluctuations, while the RMSD of the \noverall complex characterizes whether the small molecule remains stably bound within the protein active pocket. \nResults demonstrated that RMSD values for both AKT1 protein and overall complexes remained below 0.6 nm \nacross all systems, converging to stability during the later stages of the 200 ns simulation. These findings indi -\ncate that all complexes achieved dynamic equilibrium, with small molecules stably maintained within the AKT1 \nactive pocket without significant dissociation. Among the six candidate small-molecule systems, the AKT1 pro-\ntein RMSD values for cerevisterol, enoic acid and kaempferol were slightly higher than those of the other three \nmolecules, suggesting that these three complexes underwent a certain conformational optimization process dur -\ning the simulation. The AKT1 protein RMSD value of the positive-control drug capivasertib system was lower \nthan those of the cerevisterol, enoic acid and kaempferol systems, but slightly higher than those of the populo -\nside_qt and GA120 systems. This indicates that, relative to the positive-control drug, populoside_qt and GA120 \nimpart superior overall conformational dynamic stability to the protein upon binding with AKT1.RMSF is used \nto quantify the mobility-flexibility of residues across different protein regions and serves as a key indicator for \nanalyzing dynamic conformational features. As shown in Fig. 6C, obvious differences in residue flexibility regu-\nlation were observed after AKT1 bound to different active molecules. The residue regions 100-140 and 290-320 \nexhibited generally higher fluctuation amplitudes, representing highly flexible regions of the protein. Compared \nwith other ligands, kaempferol binding markedly increased the residue flexibility within the 100–140 segment. \nGA120 mainly enhanced the flexibility of residues in the 290-310 region. By contrast, baicalein exerted the opti-\nmal conformational-stabilizing effect on both flexible loops. The positive control capivasertib exerts relatively \nweak perturbation on the residue flexibility of different AKT1 domains, and its RMSF values in the 100–140 \nand 290–320 regions are both less than 0.6 nm. The central core region of the protein maintained low flexibility \nthroughout the simulation, and no overall structural disruption occurred. Hydrogen bonds act as critical driv -\ning forces for ligand-protein binding. Their number and stability constitute important parameters for evaluating \nthe binding capacity of complexes. Statistics on hydrogen-bond counts derived from simulation trajectories are \npresented in Fig. 6D. The kaempferol-AKT1 complex formed the largest number of hydrogen bonds, followed \nby Populoside_qt-AKT1 and enoic acid-AKT1 complexes. This observation can be mainly attributed to multiple \nhydroxyl or phenolic-hydroxyl groups contained in these molecules, which facilitate stable polar interactions \nwith protein residues.\nBinding free energies between the six candidate molecules and AKT1 protein were further calculated based on \nMM/GBSA. As shown in Fig. 6E, kaempferol exhibited the lowest binding free energy with AKT1 at -34.03 kcal/\nmol, followed by cerevisterol (-33.7 kcal/mol), Populoside_qt (-28.92 kcal/mol), baicalein (-24.04 kcal/mol), \nenoic acid (-22.42 kcal/mol), and GA120 (-20.13 kcal/mol). The binding free energy of the positive control was \nonly − 6.02 kcal/mol.These results indicate that kaempferol possesses stronger binding affinity to AKT1 compared \nto the other five molecules and far exceeds that of the positive control. Interactions between small molecules and \nkey residues of AKT1 protein were visualized using the ProLIF tool (Fig. 7). Hydrophobic interactions, hydrogen \nbonds, van der Waals forces, and pi-stacking interactions contributed to the specificity and stability of ligand-\nreceptor binding. Combined with molecular docking results, the binding free energy calculations from molecular \ndynamics simulations showed consistent trends. Notably, absolute value differences existed between the two \nmethods, which is expected due to fundamental methodological differences. Molecular docking represents static \nconformational calculations, whereas molecular dynamics simulations incorporate dynamic conformational \nchanges and environmental interactions. Furthermore, the structural superposition of the initial (0 ns) and final \n(200 ns) conformations for each complex system demonstrated a high degree of overlap across all six ligand–\nAKT1 complexes and the positive-control capivasertib–AKT1 complex, with no evidence of protein unfolding or \nligand dissociation from the binding pocket, directly reflecting the conformational stability of the systems (Figure \nS1). In conjunction with the RMSD and RMSF trajectory data—which showed convergent RMSD curves during \n\n\nArticle in Press\nScientific Reports\nhttps://doi.org/10.1038/s41598-026-77411-x\nthe later stages of the simulation and maintained low RMSF values in the protein core functional domains—the \nquantitative trajectory parameters and the conformational superposition visualizations mutually corroborate each \nother, collectively providing comprehensive evidence that the complexes obtained from molecular docking main-\ntained good structural stability throughout the 200 ns simulation. Collectively, molecular dynamics simulation \nresults further confirmed the stable binding capabilities between core active ingredients (particularly kaempferol \nand cerevisterol) and AKT1 from the perspectives of dynamic conformational stability and binding free energy. \nThese findings provide robust kinetic evidence for subsequent investigation into their molecular mechanisms of \nregulating AKT1-related signaling pathways.\nDiscussion\nAM and EM are classified under the categories of “Zheng Jia” (abdominal masses) and “dysmenorrhea” with \ntheir core pathogenesis often characterized by the intermingling of “blood stasis” and “water-dampness” in TCM. \nBlood stasis refers to impaired blood circulation accompanied by local stagnation, while water-dampness denotes \nabnormal fluid metabolism resulting in the accumulation of pathological products. Modern research has demon-\nstrated that the onset and progression of AM and EM are closely associated with chronic inflammation, immune \ndysregulation, aberrant angiogenesis, and extracellular matrix remodeling3,34. These contemporary findings align \nconceptually with the TCM framework of blood stasis and water-dampness. For instance, inflammatory responses \nand angiogenesis with ectopic endometrial tissues may contribute to local microcirculatory disturbances, thereby \nfostering the formation of blood stasis3. Concurrently, local inflammation and tissue injury may be accompanied \nby edema and exudation, manifesting as water-dampness retention. Although AM and EM are clinically distinct \nentities, they exhibit substantial similarities in etiology, pathogenesis, and symptomatology. Recent single-cell \nFig. 7 Interactions and \nstability between molecules \nand key residues of AKT1 \nprotein. (A) Baicalein, (B) \ncerevisterol, (C) Enoic acid, \n(D) GA120, (E) Kaempferol, \n(F) Populoside_qt.\n \n\n\nArticle in Press\nScientific Reports\nhttps://doi.org/10.1038/s41598-026-77411-x\nsequencing studies have revealed that both conditions harbor analogous pathological cell types, differing primar-\nily in their anatomical locations (intrauterine or extrauterine) 3,16. This suggests that, despite variations in lesion \nsite, their underlying cytopathological foundations are likely comparable. Both disorders are characterized by \nthe proliferation and invasive growth of ectopic endometrial tissue, inflammatory responses, and dependence on \novarian hormones. Given these shared pathogenic features, therapeutic strategies effective for one condition are \nfrequently applicable to the other. TCM approaches-including activating blood circulation to resolve stasis, clear-\ning heat and promoting diuresis, and regulating the liver, spleen, and kidney-are widely employed in managing \nboth diseases. GFW has demonstrated significant clinical efficacy in treating AM and EM by improving blood \ncirculation and alleviating symptoms35–38.\nBased on the screening of active components of GFW against AM and EM, combined with molecular dock -\ning and 200 ns molecular dynamics simulation results, cerevisterol, kaempferol and baicalein were identified \nas key candidate active constituents of this formula. Molecular docking results showed that these three com -\nponents could form favorable binding with core disease-related targets of AM and EM, and establish multiple \nnon-covalent interactions such as hydrogen bonds and hydrophobic interactions with key amino-acid residues \nwithin the binding pocket. Molecular dynamics simulations further verified that these complex systems tended to \nbe stable throughout the 200 ns simulation period, suggesting a structural basis for their binding to target proteins. \nPublished studies provide partial evidence supporting the functional properties of these three candidate compo -\nnents. As a sterol compound, cerevisterol exhibits potent anti-inflammatory activity. It can modulate inflamma -\ntory pathways including MAPK, NF-κB and Nrf2/HO-1, and reduce the release of multiple pro-inflammatory \nfactors39. Moreover, it can regulate macrophage polarization via the PI3K/AKT/mTOR pathway to alleviate \ninflammatory injury40. Kaempferol, a natural flavonoid, suppresses the release of inflammatory mediators such \nas TNF-α and IL-6, modulates inflammatory responses mediated by the MAPK signaling axis, and scavenges \nreactive oxygen species to relieve cellular oxidative-stress damage 41–43. It can also induce relaxation of uterine \nsmooth muscle, showing potential regulatory effects on abnormal intrauterine lesions 44. Baicalein derived from \nScutellaria baicalensis has been reported to inhibit proliferation and invasion of ectopic endometrial stromal cells \nand produce lesion-suppressive effects in endometriosis-related models45,46. In addition, it can ameliorate repro-\nductive-system injuries associated with oxidative stress, inflammatory disturbance and insulin resistance 47,48. \nFrom the perspective of TCM theory, “blood-stasis-dampness accumulation” represents the shared pathogenesis \nof AM and EM. Combined with our findings, it can be speculated that the three core candidate active compo -\nnents of GFW (cerevisterol, kaempferol and baicalein) may exert multi-target synergistic effects through multiple \npathways including anti-inflammation, anti-oxidation, angiogenesis regulation, immunomodulation and apop -\ntosis mediation. They may constitute a potential regulatory network of “activating blood circulation to drain \ndampness and promoting dampness elimination to facilitate blood circulation”. These findings provide partial \npharmacological hypothesis-based evidence for the mechanism by which Guizhi Fuling Wan intervenes in the \n“blood-stasis-dampness accumulation” pathogenesis and treats AM and EM.\nBased on the PPI network of GFW against AM and EM, TP53, AKT1, MAPK3, JUN, HSP90AA1, BCL2 and \nESR1 were screened as the core target cluster of the drug. Molecular docking results further suggested that AKT1, \nMAPK3 and HSP90AA1 possessed the optimal binding stability with small-molecule ligands, which were identi-\nfied as the most critical candidate target proteins responsible for the pharmacological effects of GFW in this study. \nThese targets were enriched in pathways related to cell-cycle regulation, apoptosis, cell survival, stress response \nand estrogen signaling, covering pathological processes including inflammatory response, immune homeostasis, \nas well as proliferation and invasion of ectopic endometrium. In the following section, based on our bioinformatic \nscreening and molecular docking results together with previously published evidence, we interpret the potential \nmulti-target regulatory mechanisms of GFW for the treatment of AM and EM. AKT1 is one of the top-ranked \ncore targets according to docking binding capacity. As a critical kinase of the PI3K/Akt pathway, accumulating \nevidence has demonstrated that aberrant activation of this pathway drives proliferation, invasion and apopto -\nsis resistance of ectopic endometrial stromal cells, mediates local lesion-associated inflammation and uterine \nmyometrial fibrosis, and contributes to the progression of AM and EM 49,50. Multiple molecular regulatory axes \n\n\nArticle in Press\nScientific Reports\nhttps://doi.org/10.1038/s41598-026-77411-x\ncan determine the fate of ectopic endometrial cells via the PI3K/Akt pathway. Capn7, miR-21, NLRC3 and \nLinc-ROR modulate pathway activity at the levels of proteolysis, microRNA, inflammatory receptor and long \nnon-coding RNA, and further regulate stromal cell decidualization, autophagy-apoptosis, cell migration-invasion \nand myometrial fibrosis51–55. Clinical sample studies also indicate that PI3K/Akt pathway activation is correlated \nwith pyroptosis and high expression of inflammatory factors in EM lesions 56. According to the target-inter -\naction network in the present study, AKT1 forms multi-layer regulatory modules with TP53, NFKB1, RELA, \nAR and BCL2. Combined with published mechanistic reports, hsa-miR-149-3p can target the 3′-UTR of AKT1 \nto suppress its protein expression and participate in post-transcriptional regulation of TP53 signaling, serving \nas a cross-regulatory node between AKT1 and TP5357,58. Transcription factors including AR, NFKB1, RELA \nand TP53 can bidirectionally modulate AKT1 activity through promoter binding, protein-protein interaction and \nalternative splicing59–64. Moreover, AKT1 can further activate the downstream anti-apoptotic protein BCL2 to \nimprove the survival capacity of ectopic endometrial cells 65. It is speculated in this study that GFW may target \nand inhibit the abnormal activation of AKT1, disturb the AKT1‑TP53‑BCL2 apoptotic regulatory loop and attenu-\nate apoptosis resistance of ectopic endometrial cells. Meanwhile, GFW may block the positive-feedback axis of \nAKT1-NF-κB and ameliorate the local inflammatory microenvironment of lesions. HSP90AA1 also exhibited \nfavorable binding affinity in molecular docking simulations. As a molecular chaperone protein, it stabilizes the \nconformational status of signaling proteins in multiple pathways. Reproductive-relevant studies have revealed \nthat HSP90AA1 is dynamically expressed throughout the menstrual cycle, and participates in cyclic endometrial \nproliferation-differentiation, maintenance of embryo-implantation microenvironment and proliferation regula -\ntion of uterine myometrial cells66–68. In addition, HSP90α derived from circulating extracellular vesicles of AM \npatients shows potential as a disease biomarker 69. Accordingly, we put forward a preliminary hypothesis: after \nbinding with HSP90AA1, active components of GFW may reduce the chaperone-mediated stabilization of sig -\nnal proteins associated with proliferation, inflammation and angiogenesis, and thereby cooperatively suppress \nectopic endometrial proliferation, lesion neovascularization and local inflammatory infiltration. MAPK3 repre-\nsents another target with high binding stability screened in our study, which belongs to the ERK cascade kinase \npathway. Existing literature confirms that abnormal activation of the MAPK/ERK pathway activates uterine \nsmooth muscle cells, mediates myometrial hyperplasia in AM, and participates in the invasive behavior of ectopic \nendometrium in EM. Hence, this pathway acts as a shared pathogenic pathway for both diseases70–72. As a down-\nstream substrate of MAPK-JNK signaling, transcription factor JUN transduces upstream kinase signals to trigger \ntranscription of inflammation- and proliferation-related genes 73. Consistent with our PPI-network results, tight \ninteractions exist between MAPK3 and JUN, indicating that the MAPK3‑JUN axis functions as an independent \nfunctional module within the GFW-target network. Based on current analytical outcomes, we propose that GFW \nmay target MAPK3, dampen excessive ERK-pathway activation, inhibit abnormal uterine smooth-muscle pro -\nliferation as well as ectopic endometrial invasion, and relieve local lesion inflammation. Collectively, based on \nour PPI-network analysis, pathway enrichment and molecular docking results, the AKT1‑TP53‑BCL2 apoptotic \nregulatory loop, the AKT1-NF-κB inflammatory feedback axis, the HSP90AA1 chaperone-regulatory module, \nand the MAPK3‑JUN proliferation-inflammation pathway together constitute the potential multi-target molecular \nnetwork underlying GFW intervention against AM and EM. All mechanistic inferences in this study are derived \nfrom public-dataset analysis and in-silico molecular docking simulations. This manuscript does not assert defi -\nnite upstream-downstream regulatory relationships among targets, nor the in-vivo binding effects between drug \ncomponents and targets. Further cellular-model experiments will be performed to verify the protein expression of \nthe top-ranked core targets and validate the regulatory hypotheses raised in this work.\nGO and KEGG pathway enrichment analyses further revealed that the therapeutic effects of GFW against AM \nand EM may also involve multiple potential core pathways, including focal adhesion, endocrine resistance, apop-\ntosis, HIF-1 signaling pathway, progesterone-mediated oocyte maturation, TNF signaling pathway, cell cycle, \nand estrogen signaling pathway. These pathways are highly consistent with the core pathological features of AM \nand EM lesions, such as aberrant proliferation and invasion, apoptosis resistance, hormonal disturbance, and \nchronic inflammation, providing omics-based predictive evidence for elucidating the potential multi-pathway \n\n\nArticle in Press\nScientific Reports\nhttps://doi.org/10.1038/s41598-026-77411-x\nsynergistic mechanism of GFW. Our enrichment results suggested that focal adhesion, cell cycle, and apoptosis \npathways represent key potential routes by which GFW modulates the biological behaviors of ectopic endome -\ntrium. The focal-adhesion pathway governs ectopic endometrial adhesion, EMT transition, invasion and migra -\ntion. This study predicts that GFW may target this pathway and potentially suppress the invasive progression of \nlesions. This finding is consistent with previous reports indicating that Guizhi Fuling-containing preparations \ncan down-regulate the expression of genes related to cell invasion and migration by inhibiting the AKT/GSK3β \nsignaling pathway74, implying that GFW may interfere with the FAK-associated signaling network to reduce the \nadhesion and invasion capacities of ectopic endometrium. Meanwhile, enrichment of apoptosis and cell-cycle \npathways indicates that GFW may reverse apoptosis resistance and arrest abnormal cell cycles to restrain lesion \nproliferation. Existing studies have confirmed that apoptosis evasion commonly occurs in ectopic endometrium \nof AM and EM patients 75–77. GFW can disturb abnormal cell cycles and induce apoptosis in lesion cells by \nregulating the Bcl‑2/Bax apoptotic balance and activating CHK2 and P2178,79, which provides solid support \nfor our pathway-enrichment predictions. Enrichment of endocrine resistance, estrogen signaling pathway, and \nprogesterone-mediated oocyte-maturation pathways reflects the potential capacity of GFW to maintain reproduc-\ntive-endocrine homeostasis and ameliorate disease-related reproductive impairment. Hormonal disturbance and \nendocrine resistance constitute core pathological mechanisms of AM and EM. Our study predicts that GFW may \ntarget these pathways to alleviate endocrine imbalance. Previous animal experiments demonstrated that GFW \ncould down-regulate abnormally expressed ERα in AM-model mice, suppress SGK1-mediated pathological inva-\nsion, and restore estrogen disturbance without disrupting normal estrous cycles 80. Furthermore, enrichment of \nthe progesterone-mediated oocyte-maturation pathway offers potential theoretical support for improved fertility \nprognosis after GFW treatment, which is in line with clinical observations that GFW assists in elevating preg -\nnancy rates and reducing recurrence rates 37. Moreover, HIF-1 and TNF signaling pathways were also enriched \nin our analysis, suggesting that hypoxia-driven vascular remodeling and chronic inflammatory imbalance are \nimportant potential targets for GFW-mediated therapy against AM and EM. HIF-1α-dependent hypoxia response \nand pathological angiogenesis are critical drivers of lesion progression. We predict that GFW may suppress \npathological neovascularization via modulating the HIF-1 pathway. This is consistent with earlier experimental \nresults showing that GFW reduces VEGF levels and inhibits aberrant angiogenesis in model mice 81. The TNF \nsignaling pathway mainly participates in the regulation of pelvic chronic inflammation. Multiple inflammatory \nfactors including TNF-α, IL-6 and IL-8 are highly expressed in AM and EM lesions 82,83. GFW can break the \nvicious cycle of local chronic inflammation by blocking the STAT1 pathway and lowering inflammatory-factor \nlevels84,85, further validating the reliability of our pathway-enrichment outcomes.\nTo further clarify the potential therapeutic characteristics of GFW against AM and EM, target-tissue localiza-\ntion prediction analysis was performed using the BioGPS database. Our preliminary analytical results indicated \nthat potential targets of GFW for AM and EM therapy were widely enriched in tissues and organs including the \nthyroid, hypothalamus, heart, liver, and multiple key immune cells, covering multiple physiological systems such \nas the endocrine, immune, cardiovascular, nervous and metabolic systems. These findings preliminarily suggest \nthat GFW may exert a potential multi-organ and multi-system synergistic regulatory pattern. Based on the above \npredictions, it can be tentatively speculated that GFW may remodel the overall “neuro-endocrine-immune-met -\nabolic” homeostasis by integrating multi-system regulatory pathways, thereby producing potential therapeu -\ntic effects on AM and EM. It should be noted that the multi-organ and multi-system regulatory characteristics \nobtained in this study are merely scientific hypotheses derived from bioinformatic analysis. Restricted by the \ninherent properties of the database, BioGPS only collects static gene expression data from healthy human sub -\njects without drug intervention, and cannot reflect dynamic molecular expression changes under pathological \nconditions or after GFW treatment. According to our target-distribution predictions, prominent enrichment of \ntargets was observed in immune-related cells. This preliminarily implies that GFW may suppress the prolifera -\ntion and invasion of ectopic endometrium by modulating immune-cell functions and reshaping local and sys -\ntemic immune homeostasis. Meanwhile, target enrichment in neuroendocrine tissues including the hypothalamus \nand thyroid suggests that GFW may interfere with the progression of hormone-dependent ectopic lesions by \n\n\nArticle in Press\nScientific Reports\nhttps://doi.org/10.1038/s41598-026-77411-x\nregulating neuroendocrine homeostasis. In addition, the target distribution in the liver, cardiovascular and ner -\nvous tissues provides systems-level evidence supporting the traditional TCM effects of GFW, such as activating \nblood circulation to resolve stasis, improving the pelvic microenvironment, and alleviating chronic pelvic pain \nand dysmenorrhea caused by AM and EM. These observations well reflect the holistic therapeutic features of \nherbal compound prescriptions.\nRecent studies on natural-product-derived drugs have demonstrated that plant-derived small-molecule com -\npounds possess diverse pharmacophores and can simultaneously modulate multiple disease-associated signal -\ning pathways, making them promising lead molecules for complex multifactorial diseases 86. Different from the \nsingle-target intervention mode of conventional chemical drugs, plant-derived natural constituents can systemati-\ncally regulate disease-related molecular networks at multiple sites, which better matches pathological disorders \ninduced by multiple contributing factors. Existing studies on the material basis and quality markers of GFW have \nconfirmed that this herbal formula contains abundant active chemical constituents, and multiple components syn-\nergistically participate in multi-dimensional disease regulation87,88. Notably, conventional ADME screening strat-\negies based on OB and DL parameters have inherent limitations. Strict cut-off thresholds tend to exclude some \nbioactive constituents with limited oral bioavailability, even though their definite pharmacological activities have \nbeen validated by in-vitro and in-vivo experiments. Cellular and animal experiments have verified that paeonol \nfrom GFW can markedly attenuate the migration and invasion of ectopic endometrial stromal cells by inhibit -\ning the HIF-1α-mediated autophagy pathway89. Amygdalin can target and suppress the Wnt/β‑catenin signaling \npathway, effectively blocking proliferation, migration and invasion of ectopic endometrial cells in endometrio -\nsis-model mice90. These components exert obvious anti-ectopic-lesion effects, yet they were eliminated during \nour standardized ADME screening workflow owing to their generally low oral bioavailability. Accordingly, the \ncore small-molecule constituents screened in this study only represent components of GFW with relatively high \nblood-absorption potential and favorable drug-likeness properties, and cannot cover all pharmacologically effec-\ntive substances of the formula, indicating certain limitations in our active-component screening scope. Mean -\nwhile, mechanistic investigations of natural products based on network pharmacology and molecular simulation \nstill face bottlenecks in clinical translation. Plant-derived small molecules frequently suffer from poor water \nsolubility and weak metabolic stability in-vivo, which may lead to discrepancies between predicted bioactivity \nfrom computational analysis and actual therapeutic effects in living organisms. Furthermore, in-silico simulation \ncannot fully clarify the in-vivo interactions, dose-effect relationships and tissue-distribution profiles of multiple \nformula components. The regulatory mechanisms deduced in this work are merely theoretical hypotheses without \nin-vivo experimental evidence, and cannot provide direct guidance for clinical practice.\nDespite the systematic bioinformatics prediction and molecular simulation analyses conducted in this study, \nseveral inherent limitations of in silico predictive research must be clearly acknowledged.First, all target and \npathway data were derived from public databases, which contain inherent heterogeneity, incomplete annotation, \nand context-independent gene expression information. The present screening results only reflect computation -\nally predicted associations rather than disease-specific, treatment-induced molecular changes in vivo. Therefore, \nall candidate targets and pathways identified in this study remain theoretical predictions rather than confirmed \ntherapeutic mechanisms.Second, molecular docking and molecular dynamics simulations only evaluate the bind-\ning stability between small-molecule compounds and target proteins at the structural level. These computational \napproaches cannot determine the functional regulatory effects (activation or inhibition) of GFW components on \ncore targets such as AKT1 and MAPK3. Furthermore, computational binding affinity cannot fully represent actual \npharmacological activity or biological function.Third, consistent with the features of network pharmacology-\nbased research, the ADME screening criteria used in this study may exclude potentially active compounds with \nlow oral bioavailability but definite pharmacological effects. Therefore, the identified core components only \nrepresent a partial material basis of GFW, which cannot fully reflect the complete pharmacological spectrum \nof this compound prescription.Fourth, this study lacks in vitro and in vivo experimental validation. No protein \nexpression detection, pathway activity verification, or functional phenotypic assays were performed. All mecha-\nnistic hypotheses regarding the anti-inflammatory, anti-proliferative, and immunomodulatory effects of GFW \n\n\nArticle in Press\nScientific Reports\nhttps://doi.org/10.1038/s41598-026-77411-x\nare purely computational inferences without empirical evidence. In addition, this study cannot clarify the dose–\neffect relationship, metabolic transformation, and tissue distribution characteristics of GFW components in living \norganisms.Finally, although multi-system regulatory characteristics were predicted based on BioGPS data, such \nresults were obtained from static gene expression profiles of healthy populations and cannot reflect the dynamic \npathological microenvironment of AM and EM lesions.To address the above limitations, multi-level in vitro and \nin vivo functional experiments will be carried out in our subsequent studies to improve the mechanistic under -\nstanding of GFW and promote the translational application of these findings.Specifically, AM- and EM-related \ncellular models will first be established to verify the in vitro pharmacological effects of the core active constitu-\nents. Target-intervention experiments focusing on the key AKT1 pathway, combined with Western blot analysis \nof core proteins and their phosphorylated expression levels, will be performed to clarify the precise regulatory \nmode (activation or inhibition) of these small molecules on the targets, thereby compensating for the limitation \nthat molecular simulation cannot determine the direction of regulatory effects.In parallel, in vivo pharmacologi-\ncal evaluation will be conducted using AM and EM animal models. Techniques including immunohistochemistry \nand protein quantification will be applied to systematically verify the regulatory effects of GFW on core pathways \nand target networks at the cellular, tissue, and whole-animal levels.This series of validation experiments will \neffectively correct the biases of bioinformatic predictions, clarify the true pharmacological mechanisms of GFW \nagainst AM and EM, and provide reliable experimental evidence for its precise clinical application.\nConclusion\nIn this study, bioinformatic analyses were performed to preliminarily predict and systematically dissect the poten-\ntial multi-target regulatory network of GFW for the treatment of AM and EM. Core candidate active constituents \nincluding cerevisterol, kaempferol and baicalein were screened out. The results preliminarily suggest that these \ncomponents may synergistically act on key target proteins such as AKT1, MAPK3, HSP90AA1, TP53, JUN, BCL2 \nand ESR1. They may modulate signaling pathways including PI3K/Akt, MAPK and focal adhesion, and exert \npotential therapeutic effects against AM and EM by interfering with common pathological processes such as \ncell proliferation, invasion, apoptosis resistance, inflammatory response and angiogenesis.Our prediction further \nindicates that GFW possesses potential cross-system synergistic regulatory characteristics. Its target molecules \nare not restricted to the uterus, but are also widely enriched in tissues and organs including the thyroid, hypo -\nthalamus, heart, liver and key immune cells, potentially participating in biological regulation of the endocrine, \nimmune, metabolic, nervous and cardiovascular systems.In summary, from six dimensions including theoretical \nbasis, candidate active components, predicted core targets, transcriptional regulatory features, critical pathways \nand tissue distribution, this study preliminarily illustrates the potential synergistic therapeutic mechanism of \nGFW featured by “multi-component-multi-target-multi-pathway-multi-system”. The findings of this work are \nmerely predictive evidence obtained from in-silico analysis. They theoretically support the scientific connotation \nof the holistic concept of traditional Chinese medicine, provide a systems-biology reference for herbal compound \nprescriptions in the therapy of complex gynecological diseases, and offer potential research clues for subsequent \nmechanistic validation and optimization of clinical strategies.\nSupplementary Information  The online version contains supplementary material available at https://doi.org/10.1038/\ns41598-026-77411-x.\nAcknowledgements Not applicable.\nAuthor contributions Jingyi Wu: Data curation, Investigation, Software, Writing-original draft. Xinting Li: Data curation, \nInvestigation, Writing-original draft. Shuikun Li: Investigation, Data curation, Visualization. Yanting Luo: Investigation, \nWriting-original draft. Ying Gao: Software, Writing-review & editing. Yanfei Tao: Methodology, Writing-original draft. \nJiangying Yang: Conceptualization, Methodology, Writing-review & editing. Weiyong Hong: Conceptualization, Writing-\nreview & editing, Supervision.\n\n\nArticle in Press\nScientific Reports\nhttps://doi.org/10.1038/s41598-026-77411-x\nFunding This work was supported by Science and Technology Plan Project of Taizhou (Grant No. 26ywb74), Medical Sci-\nence and Technology Project of Zhejiang Province, (Grant No. 2024KY536).\nData availability The datasets generated and/or analyzed during the current study are available from the following public \nrepositories: the raw gene expression data from GEO under accession numbers GSE193928, GSE190580, GSE185392, \nGSE212787, GSE192636, GSE51981, and GSE6364 ( https://www.ncbi.nlm.nih.gov/geo/); the chemical component data \nfrom TCMSP ( https://tcmsp-e.com/); target prediction data from SwissTargetPrediction  ( h t t p : / / w w w . s w i s s t a r g e t p r e d i c t i o \nn . c h / ) ; protein–protein interaction data from STRING ( https://string-db.org/). All other data generated or analyzed during \nthis study are included in the published article and its supplementary information files. The molecular dynamics simulation \ntrajectories and analysis scripts are available from the corresponding author upon reasonable request.\nDeclarations\nCompeting interests The authors declare no competing interests.\nOpen Access  This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 Interna-\ntional License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as \nlong as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, \nand indicate if you modified the licensed material. You do not have permission under this licence to share adapted material \nderived from this article or parts of it. The images or other third party material in this article are included in the article’s Cre-\native Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s \nCreative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you \nwill need to obtain permission directly from the copyright holder. To view a copy of this licence, visit  h t t p : / / c r e a t i v e c o m m o \nn s . o r g / l i c e n s e s / b y - n c - n d / 4 . 0 /     .  \nReferences\n 1. Palumbo, M. et al. Genetic and epigenetic components in the pathogenesis of adenomyosis and endometriosis in \nadolescents. Biomedicines 13, 2988 (2025).\n 2. Bordonné, C. et al. Imaging for evaluation of endometriosis and adenomyosis. Minerva Obstet. Gynecol. 73, 290–303. \nhttps://doi.org/10.23736/s2724-606x.21.04710-9 (2021).\n 3. Giudice, L. C., Liu, B. & Irwin, J. C. Endometriosis and adenomyosis unveiled through single-cell glasses. Am. J. Obstet. \nGynecol. 232, S105–s123. https://doi.org/10.1016/j.ajog.2024.08.043 (2025).\n 4. Manoukian, S. B., Shaheen, N. H. & Kowal, D. J. Gynecologic adenomyosis and endometriosis: Key imaging findings, \nmimics, and complications. Contemp. Diagn. Radiol. 38, 1–6 (2015).\n 5. Deryabina, V . A., Brus, T. V . & Pyurveev, S. S. Endometriosis and adenomyosis: Common and disparate features. \nPediatrician (St. Petersburg) https://doi.org/10.17816/PED13581-97 (2022).\n 6. Maruyama, S., Imanaka, S., Nagayasu, M., Kimura, M. & Kobayashi, H. Relationship between adenomyosis and \nendometriosis; Different phenotypes of a single disease? Eur. J. Obstet. Gynecol. Reprod. Biol. 253, 191–197.  h t t p s : / / d o \ni . o r g / 1 0 . 1 0 1 6 / j . e j o g r b . 2 0 2 0 . 0 8 . 0 1 9     (2020).\n 7. Petraglia, F., Vannuccini, S., Santulli, P., Marcellin, L. & Chapron, C. An update for endometriosis management: A \nposition statement. J. Endometr. Uterine Disord. 6, 100062. https://doi.org/10.1016/j.jeud.2024.100062 (2024).\n 8. Habiba, M., Guo, S.-W. & Benagiano, G. Are adenomyosis and endometriosis phenotypes of the same disease process?. \nBiomolecules 14, 32 (2024).\n 9. Donnez, J., Stratopoulou, C. A. & Dolmans, M. M. Endometriosis and adenomyosis: Similarities and differences. Best \nPract. Res. Clin. Obstet. Gynaecol. 92, 102432. https://doi.org/10.1016/j.bpobgyn.2023.102432 (2024).\n 10. Benagiano, G. & Brosens, I. Adenomyosis and endometriosis have a common origin.\n 11. Bulun, S. E. et al. Endometriosis and adenomyosis: shared pathophysiology. Fertil. Steril. 119, 746–750.  h t t p s : / / d o i . o r g / \n1 0 . 1 0 1 6 / j . f e r t n s t e r t . 2 0 2 3 . 0 3 . 0 0 6     (2023).\n 12. Gabidullina, R. I. et al. Adenomyosis and endometriosis: Common features. Pract. Med.  h t t p s : / / d o i . o r g / 1 0 . 3 2 0 0 0 / 2 0 7 2 - 1 \n7 5 7 - 2 0 2 3 - 4 - 3 7 - 4 0     (2023).\n 13. Di Donato, N. & Seracchioli, R. How to evaluate adenomyosis in patients affected by endometriosis?. Minim. Invasive \nSurg. 2014, 507230. https://doi.org/10.1155/2014/507230 (2014).\n 14. Scala, C. et al. Influence of adenomyosis on pregnancy and perinatal outcomes in women with endometriosis. Ultrasound \nObstet. Gynecol. 52, 666–671. https://doi.org/10.1002/uog.18989 (2018).\n\n\nArticle in Press\nScientific Reports\nhttps://doi.org/10.1038/s41598-026-77411-x\n 15. Gonzales, M. et al. Patients with adenomyosis are more likely to have deep endometriosis. Gynecol. Surg. 9, 259–264. \nhttps://doi.org/10.1007/s10397-012-0746-4 (2012).\n 16. Rytkönen, K. T., Zúñiga Norman, S., Elo, L. L. & Poutanen, M. P-335 Integrative analysis of endometriosis and \nadenomyosis through single-cell transcriptomics. Hum. Reprod. https://doi.org/10.1093/humrep/deae108.276 (2024).\n 17. Li, H. & Zhu, Y . The chemical composition analysis of Yixin Tongmai Granules using UHPLC-MS/MS and exploration \nof its potential mechanism in treatment of coronary artery disease based on network pharmacology and molecular \ndocking. Medicine (Baltimore) 104, e41620. https://doi.org/10.1097/MD.0000000000041620 (2025).\n 18. Wu, W. et al. Mechanisms underlying the therapeutic effects of Gang Huo Qing wen granules in the treatment of influenza \nbased on network pharmacology, molecular docking and molecular dynamics. Sci. Rep. 14, 15853.  h t t p s : / / d o i . o r g / 1 0 . 1 0 \n3 8 / s 4 1 5 9 8 - 0 2 4 - 6 2 4 6 9 - 2     (2024).\n 19. Dilawari, A. et al. US Food and Drug Administration approval summary: Capivasertib with fulvestrant for hormone \nreceptor-positive, human epidermal growth factor receptor 2-negative locally advanced or metastatic breast cancer with \nPIK3CA/AKT1/PTEN alterations. J. Clin. Oncol. 42, 4103–4113. https://doi.org/10.1200/JCO.24.00427 (2024).\n 20. Cao, K. et al. Discovery of AKT1 inhibitors for obesity and metabolic dysfunction-associated steatotic liver disease using \nQSAR-guided virtual screening and Gaussian accelerated molecular dynamics. ACS Omega 11, 20461–20479.  h t t p s : / / d \no i . o r g / 1 0 . 1 0 2 1 / a c s o m e g a . 5 c 1 1 6 8 8     (2026).\n 21. Uhlenbrock, N. et al. Structural and chemical insights into the covalent-allosteric inhibition of the protein kinase Akt. \nChem. Sci. 10, 3573–3585. https://doi.org/10.1039/c8sc05212c (2019).\n 22. Cheng, R. et al. High‐resolution crystal structure of human Mapkap kinase 3 in complex with a high affinity ligand. \nProtein Sci. 19, 168–173. https://doi.org/10.1002/pro.294 (2009).\n 23. Wright, L. et al. Structure-activity relationships in purine-based inhibitor binding to HSP90 isoforms. Chem. Biol. 11, \n775–785. https://doi.org/10.1016/j.chembiol.2004.03.033 (2004).\n 24. Balourdas, D. I., Markl, A. M., Kramer, A., Settanni, G. & Joerger, A. C. Structural basis of p53 inactivation by cavity-\ncreating cancer mutations and its implications for the development of mutant p53 reactivators. Cell Death Dis. 15, 408. \nhttps://doi.org/10.1038/s41419-024-06739-x (2024).\n 25. Hong, S. et al. Methyl-dependent and spatial-specific DNA recognition by the orthologous transcription factors human \nAP-1 and Epstein-Barr virus Zta. Nucleic Acids Res. 45, 2503–2515. https://doi.org/10.1093/nar/gkx057 (2017).\n 26. Guo, Y . et al. Discovery of the clinical candidate Sonrotoclax (BGB-11417), a highly potent and selective inhibitor \nfor both WT and G101V mutant Bcl–2. J. Med. Chem. 67, 7836–7858. https://doi.org/10.1021/acs.jmedchem.4c00027 \n(2024).\n 27. Nwachukwu, J. C. et al. Systems structural biology analysis of ligand effects on ERalpha predicts cellular response to \nenvironmental estrogens and anti-hormone therapies. Cell. Chem. Biol. 24, 35–45.  h t t p s : / / d o i . o r g / 1 0 . 1 0 1 6 / j . c h e m b i o l . 2 0 \n1 6 . 1 1 . 0 1 4     (2017).\n 28. Huang, J., Gao, J. & Chen, Q. An interpretable deep learning and molecular docking framework for repurposing existing \ndrugs as inhibitors of SARS-CoV-2 main protease. Molecules https://doi.org/10.3390/molecules30163409 (2025).\n 29. Park, S. J., Borin, B. N., Martinez-Yamout, M. A. & Dyson, H. J. The client protein p53 adopts a molten globule-like state \nin the presence of Hsp90. Nat. Struct. Mol. Biol. 18, 537–541. https://doi.org/10.1038/nsmb.2045 (2011).\n 30. Li, X. et al. Proteomic analyses reveal distinct chromatin-associated and soluble transcription factor complexes. Mol. \nSyst. Biol. https://doi.org/10.15252/msb.20145504 (2015).\n 31. Mo, X. et al. Systematic discovery of mutation-directed neo-protein–protein interactions in cancer. Cell 185, 1974–\n1985e1912. https://doi.org/10.1016/j.cell.2022.04.014 (2022).\n 32. Mancini, F. et al. MDM4 (MDMX) localizes at the mitochondria and facilitates the p53-mediated intrinsic-apoptotic \npathway. EMBO J. 28, 1926–1939. https://doi.org/10.1038/emboj.2009.154 (2009).\n 33. Kanehisa, M., Furumichi, M., Sato, Y ., Matsuura, Y . & Ishiguro-Watanabe, M. KEGG: Biological systems database as a \nmodel of the real world. Nucleic Acids Res. 53, D672–D677. https://doi.org/10.1093/nar/gkae909 (2025).\n 34. Mitra, I. et al. MMP2/TIMP2 ratio as a measurable indicator for differentiating diffuse adenomyosis, ovarian endometriosis, \nand their co-existence. Eur. J. Obstet. Gynecol. Reprod. Biol. 314, 114669 (2025).\n 35. Liu, T., Ma, S. & Si, X. Acupuncture combined with Guizhi Fuling Wan decoction for adenomyosis-associated \ndysmenorrhea: A case report. Explore (NY) 21, 103169. https://doi.org/10.1016/j.explore.2025.103169 (2025).\n 36. Qin, Y . et al. Meta-analysis and network pharmacology studies of the clinical efficacy of Guizhi Fuling capsules/pills \ncombined with dienogest in treating endometriosis. Medicine (Baltimore) 103, e40528.  h t t p s : / / d o i . o r g / 1 0 . 1 0 9 7 / m d . 0 0 0 0 \n0 0 0 0 0 0 0 4 0 5 2 8     (2024).\n 37. Meng, W., Ta, N. & Wang, F. Add-on effect of Guizhi Fuling formula to mifepristone for endometriosis: A meta-analysis \nof randomized controlled trials. Medicine (Baltimore)  98, e16878. https://doi.org/10.1097/md.0000000000016878 \n(2019).\n 38. Armour, M. et al. The effectiveness of a modified Gui Zhi Fu Ling Wan formulation (Gynoclear™) for the treatment \nof endometriosis: a study protocol for a placebo-controlled, double-blind, randomised controlled trial. Trials 22, 299. \nhttps://doi.org/10.1186/s13063-021-05265-x (2021).\n\n\nArticle in Press\nScientific Reports\nhttps://doi.org/10.1038/s41598-026-77411-x\n 39. Alam, M. B. et al. Cerevisterol alleviates inflammation via suppression of MAPK/NF-κB/AP-1 and activation of the \nNrf2/HO-1 signaling cascade. Biomolecules https://doi.org/10.3390/biom10020199 (2020).\n 40. Ma, X. N. et al. Validating the gouty arthritis alleviating effects of cerevisterol through integrated in silico, in vitro, and \nin vivo studies. In Proceedings of the ACR Convergence (2025).\n 41. Alam, W., Khan, H., Shah, M. A., Cauli, O. & Saso, L. Kaempferol as a dietary anti-inflammatory agent: Current \ntherapeutic standing. Molecules https://doi.org/10.3390/molecules25184073 (2020).\n 42. Qin, P. & Wu, H. Mechanism of Kaempferol in anti-inflammation based on network pharmacology. IOP Conf. Ser.: Earth \nEnviron. Sci. 692, 032031. https://doi.org/10.1088/1755-1315/692/3/032031 (2021).\n 43. Shahbaz, M. et al. Anticancer, antioxidant, ameliorative and therapeutic properties of kaempferol. Int. J. Food Prop. 26, \n1140–1166. https://doi.org/10.1080/10942912.2023.2205040 (2023).\n 44. Revuelta, M. P., Cantabrana, B. & Hidalgo, A. Mechanisms involved in kaempferol-induced relaxation in rat uterine \nsmooth muscle. Life Sci. 67, 251–259. https://doi.org/10.1016/S0024-3205(00)00627-5 (2000).\n 45. Jin, Z., Huang, J. & Zhu, Z. Baicalein reduces endometriosis by suppressing the viability of human endometrial stromal \ncells through the nuclear factor-κB pathway in vitro. Exp. Ther. Med. 14, 2992–2998.  h t t p s : / / d o i . o r g / 1 0 . 3 8 9 2 / e t m . 2 0 1 7 . \n4 8 6 0     (2017).\n 46. Ke, J. Y . et al. Baicalein inhibits FURIN-MT1-MMP-mediated invasion of ectopic endometrial stromal cells in \nendometriosis possibly by reducing the secretion of TGFB1. Am. J. Reprod. Immunol. 85, e13344.  h t t p s : / / d o i . o r g / 1 0 . 1 1 1 \n1 / a j i . 1 3 3 4 4     (2021).\n 47. Li, Y . Y . et al. Baicalein improves the symptoms of polycystic ovary syndrome by mitigating oxidative stress and \nferroptosis in the ovary and gravid placenta. Phytomedicine 128, 155423. https://doi.org/10.1016/j.phymed.2024.155423 \n(2024).\n 48. Yang, Z., Huang, W., Zhang, J., Xie, M. & Wang, X. Baicalein improves glucose metabolism in insulin resistant HepG2 \ncells. Eur. J. Pharmacol. 854, 187–193. https://doi.org/10.1016/j.ejphar.2019.04.005 (2019).\n 49. Driva, T. S., Schatz, C. & Haybaeck, J. Endometriosis-associated ovarian carcinomas: How PI3K/AKT/mTOR pathway \naffects their pathogenesis. Biomolecules https://doi.org/10.3390/biom13081253 (2023).\n 50. Ma, T. et al. Imperatorin alleviated endometriosis by inhibiting the activation of PI3K/Akt/NF-κB pathway in rats. Life \nSci. 274, 119291. https://doi.org/10.1016/j.lfs.2021.119291 (2021).\n 51. Kang, N. et al. Calpain7 negatively regulates human endometrial stromal cell decidualization in EMs by promoting \nFoxO1 nuclear exclusion via hydrolyzing AKT1. Biol. Reprod. 106, 1112–1125. https://doi.org/10.1093/biolre/ioac041 \n(2022).\n 52. Jia, J. J., Lai, H. J., Sun, B. W., Lu, J. & Zeng, Y . Y . MiR-21 regulates autophagy and apoptosis of ectopic endometrial \nstromal cells of adenomyosis via PI3K/AKT/mTOR pathway. Sci. Rep. 157639.  h t t p s : / / d o i . o r g / 1 0 . 1 0 3 8 / s 4 1 5 9 8 - 0 2 5 - 9 2 \n5 2 6 - 3     (2025).\n 53. Yang, Q. et al. Integrated multi-omics analysis and experimental verification reveal the involvement of the PI3K/Akt \nsignaling pathway in myometrial fibrosis of adenomyosis. Sci. Rep. 15, 13637.  h t t p s : / / d o i . o r g / 1 0 . 1 0 3 8 / s 4 1 5 9 8 - 0 2 5 - 9 8 3 \n6 9 - 2     (2025).\n 54. Jiang, Z. et al. NLRC3 inhibits the migration and invasion of adenomyosis by modulating the PI3K/AKT/mTOR pathway \nin endometrial cells. Crit. Rev. Immunol. 45, 13–28. https://doi.org/10.1615/CritRevImmunol.2025059822 (2025).\n 55. Xu, X. Y . et al. Linc-ROR promotes endometrial cell proliferation by activating the PI3K-Akt pathway. Eur. Rev. Med. \nPharmacol. Sci. 22, 2218–2225. https://doi.org/10.26355/eurrev_201804_14807 (2018).\n 56. An, M. et al. PI3K/AKT signaling pathway associates with pyroptosis and inflammation in patients with endometriosis. \nJ. Reprod. Immunol. 162, 104213. https://doi.org/10.1016/j.jri.2024.104213 (2024).\n 57. Si, L. et al. Potent effects of dioscin against pancreatic cancer via miR-149-3P-mediated inhibition of the Akt1 signalling \npathway. Br. J. Pharmacol. 174, 553–568. https://doi.org/10.1111/bph.13718 (2017).\n 58. Zou, A. et al. LINC00472 acts as a tumor suppressor in NSCLC through KLLN-mediated p53-Signaling Pathway via \nMicroRNA-149-3p and MicroRNA-4270. Mol. Ther. Nucleic Acids 17, 563–577.  h t t p s : / / d o i . o r g / 1 0 . 1 0 1 6 / j . o m t n . 2 0 1 9 . 0 \n6 . 0 0 3     (2019).\n 59. Shah, K. et al. Androgen receptor signaling regulates the transcriptome of prostate cancer cells by modulating global \nalternative splicing. Oncogene 39, 6172–6189. https://doi.org/10.1038/s41388-020-01429-2 (2020).\n 60. Meng, F. & D’Mello, S. R. NF-kappaB stimulates Akt phosphorylation and gene expression by distinct signaling \nmechanisms. Biochim. Biophys. Acta. 1630, 35–40. https://doi.org/10.1016/j.bbaexp.2003.09.001 (2003).\n 61. Zhou, W. et al. The AKT1/NF-kappaB/Notch1/PTEN axis has an important role in chemoresistance of gastric cancer \ncells. Cell Death Dis. 4, e847. https://doi.org/10.1038/cddis.2013.375 (2013).\n 62. Kwon, O. et al. Complex formation of p65/RelA with nuclear Akt1 for enhanced transcriptional activation of NF-kappaB. \nBiochem. Biophys. Res. Commun. 365 (4), 771–776 (2008).\n 63. Feng, Z. et al. The regulation of AMPK beta1, TSC2, and PTEN expression by p53: stress, cell and tissue specificity, and \nthe role of these gene products in modulating the IGF-1-AKT-mTOR pathways. Cancer Res. 67, 3043–3053.  h t t p s : / / d o i \n. o r g / 1 0 . 1 1 5 8 / 0 0 0 8 - 5 4 7 2 . C a n - 0 6 - 4 1 4 9     (2007).\n\n\nArticle in Press\nScientific Reports\nhttps://doi.org/10.1038/s41598-026-77411-x\n 64. Chen, M. et al. A p53-phosphoinositide signalosome regulates nuclear AKT activation. Nat. Cell. Biol. 24, 1099–1113. \nhttps://doi.org/10.1038/s41556-022-00949-1 (2022).\n 65. Su, H., Peng, C. & Liu, Y . Regulation of ferroptosis by PI3K/Akt signaling pathway: A promising therapeutic axis in \ncancer. Front. Cell. Dev. Biol. 12, 1372330. https://doi.org/10.3389/fcell.2024.1372330 (2024).\n 66. Komatsu, T. et al. Messenger ribonucleic acid expression of heat shock proteins HSP70 and HSP90 in human endometrium \nand myometrium during the menstrual cycle. J. Clin. Endocrinol. Metab. 82, 1385–1389.  h t t p s : / / d o i . o r g / 1 0 . 1 2 1 0 / j c e m . 8 \n2 . 5 . 3 9 2 7     (1997).\n 67. Tabibzadeh, S. & Broome, J. Heat shock proteins in human endometrium throughout the menstrual cycle. Infect. Dis. \nObstet. Gynecol. 7, 5–9. https://doi.org/10.1155/s1064744999000022 (1999).\n 68. Bhatti, M., Dinn, S. E., Miśkiewicz, E. & MacPhee, D. J. Expression of heat shock factor 1, heat shock protein 90 and \nassociated signaling proteins in pregnant rat myometrium: Implications for myometrial proliferation. Reprod. Biol. 19(4), \n374–85 (2019).\n 69. Chen, D. et al. Comparative proteomics identify HSP90A, STIP1 and TAGLN-2 in serum extracellular vesicles as potential \ncirculating biomarkers for human adenomyosis. Exp. Ther. Med. 23, 374. https://doi.org/10.3892/etm.2022.11301 (2022).\n 70. Makker, A., Goel, M. M., Das, V . & Agarwal, A. PI3K-Akt-mTOR and MAPK signaling pathways in polycystic ovarian \nsyndrome, uterine leiomyomas and endometriosis: An update. Gynecol. Endocrinol. 28, 175–181.  h t t p s : / / d o i . o r g / 1 0 . 3 1 0 \n9 / 0 9 5 1 3 5 9 0 . 2 0 1 1 . 5 8 3 9 5 5     (2012).\n 71. Bora, G. & Yaba, A. The role of mitogen-activated protein kinase signaling pathway in endometriosis. J. Obstet. Gynaecol. \nRes. 47, 1610–1623. https://doi.org/10.1111/jog.14710 (2021).\n 72. Streuli, I., Santulli, P., Chouzenoux, S., Chapron, C. & Batteux, F. Activation of the MAPK/ERK cell-signaling pathway \nin uterine smooth muscle cells of women with adenomyosis. Reprod. Sci. 22, 1549–1560.  h t t p s : / / d o i . o r g / 1 0 . 1 1 7 7 / 1 9 3 3 7 \n1 9 1 1 5 5 8 9 4 1 0     (2015).\n 73. Yan, H., He, L., Lv, D., Yang, J. & Yuan, Z. The role of the dysregulated JNK signaling pathway in the pathogenesis of \nhuman diseases and its potential therapeutic strategies: A comprehensive review. Biomolecules  h t t p s : / / d o i . o r g / 1 0 . 3 3 9 0 / b \ni o m 1 4 0 2 0 2 4 3     (2024).\n 74. Tao, F. et al. Fuling granule, a traditional Chinese medicine compound, suppresses cell proliferation and TGFβ-induced \nEMT in ovarian cancer. PLoS One 11, e0168892. https://doi.org/10.1371/journal.pone.0168892 (2016).\n 75. d’Argent, M. et al. Are lower levels of apoptosis and autophagy behind adenomyotic lesion survival?. Reprod. Biomed. \nOnline 47103248. https://doi.org/10.1016/j.rbmo.2023.06.003 (2023).\n 76. Toniyan, K. A. et al. Cytochrome-c-oxidase and ATPsynthase content increases in the endometrium of the patients with \nadenomyosis. Mol. Biol. Rep. 50, 3919–3925. https://doi.org/10.1007/s11033-023-08269-9 (2023).\n 77. Kobayashi, H., Imanaka, S., Yoshimoto, C., Matsubara, S. & Shigetomi, H. Molecular mechanism of autophagy and \napoptosis in endometriosis: Current understanding and future research directions. Reprod. Med. Biol. 23, e12577.  h t t p s : \n/ / d o i . o r g / 1 0 . 1 0 0 2 / r m b 2 . 1 2 5 7 7     (2024).\n 78. Hu, C. et al. Guizhi fuling capsule, an ancient Chinese formula, attenuates endometriosis in rats via induction of apoptosis. \nClimacteric 17, 410–416. https://doi.org/10.3109/13697137.2013.876618 (2014).\n 79. Lu, C. C. et al. The investigation of a traditional Chinese medicine, Guizhi Fuling Wan (GFW) as an intravesical \ntherapeutic agent for urothelial carcinoma of the bladder. BMC Complement. Altern. Med. 13, 44.  h t t p s : / / d o i . o r g / 1 0 . 1 1 8 \n6 / 1 4 7 2 - 6 8 8 2 - 1 3 - 4 4     (2013).\n 80. Wang, H., Zhang, J., Zhu, Q., Fu, X. & Li, C. Integrating network pharmacology and experimental validation deciphers \nthe mechanism of Guizhi fuling wan against adenomyosis. Evid. Based Complement. Alternat. Med.  h t t p s : / / d o i . o r g / 1 0 . 1 \n1 5 5 / 2 0 2 1 / 6 0 3 4 1 4 7     (2021).\n 81. Shi, Y . et al. Analysis of the mechanism of GuizhiFuling wan in treating adenomyosis based on network pharmacology \ncombined with molecular docking and experimental verification. Evid. Based Complement. Alternat. Med.  h t t p s : / / d o i . o r \ng / 1 0 . 1 1 5 5 / 2 0 2 2 / 6 3 5 0 2 5 7     (2022).\n 82. Bourdon, M. et al. Immunological changes associated with adenomyosis: a systematic review. Hum. Reprod. Update. 27, \n108–129. https://doi.org/10.1093/humupd/dmaa038 (2021).\n 83. Zhang, H. et al. Immune and endocrine regulation in endometriosis: What we know. J. Endometr. Uterine Disord. 4, \n100049. https://doi.org/10.1016/j.jeud.2023.100049 (2023).\n 84. Jeong, S. J. et al. Anti-inflammatory actions of herbal formula Gyejibokryeong-Hwan regulated by inhibiting chemokine \nproduction and STAT1 activation in HaCaT cells. Biol. Pharm. Bull. 38, 425–434.  h t t p s : / / d o i . o r g / 1 0 . 1 2 4 8 / b p b . b 1 4 - 0 0 6 6 \n0     (2015).\n 85. Fu, J. et al. Guizhi Fuling capsules can alleviate bortezomib-induced peripheral neuropathy by decreasing interleukin-6 \nlevels to regulate mTOR pathway-induced autophagy. Phytomedicine 139, 156494.  h t t p s : / / d o i . o r g / 1 0 . 1 0 1 6 / j . p h y m e d . 2 0 \n2 5 . 1 5 6 4 9 4     (2025).\n 86. Nare, S. & Pote, S. Plant-derived molecules in network-based drug discovery. J. Integr. Transl. Biomed.  h t t p s : / / d o i . o r g / \n1 0 . 6 7 3 1 6 / j i n t b . v 1 . i 1 . 2     (2026).\n 87. Chen, J. et al. Research on quality markers of Guizhi Fuling prescription for endometriosis treatment based on gray \ncorrelation analysis strategy. Front. Pharmacol. 11, 588549. https://doi.org/10.3389/fphar.2020.588549 (2020).\n\n\nArticle in Press\nScientific Reports\nhttps://doi.org/10.1038/s41598-026-77411-x\n 88. Cao, Z. et al. Study on contribution of main components in Guizhi Fuling capsule based on molecular imprinting \ntechnique and activity screening. Zhongguo Zhong Yao Za Zhi 40, 2420–2427. https://doi.org/10.4268/cjcmm20151230 \n(2015).\n 89. Pang, C. et al. Paeonol alleviates migration and invasion of endometrial stromal cells by reducing HIF-1α-regulated \nautophagy in endometriosis. Front. Biosci. (Landmark Ed). 26, 485–495. https://doi.org/10.52586/4961 (2021).\n 90. Yu, M., Yang, L., Pei, Y . & Xu, M. Amygdalin inhibits endometrial stromal cell proliferation, migration, and invasion in \nendometriosis mice via inhibiting Wnt/β-catenin signaling. J. Mol. Histol. 56, 11.  h t t p s : / / d o i . o r g / 1 0 . 1 0 0 7 / s 1 0 7 3 5 - 0 2 4 - 1 \n0 3 0 1 - 6     (2024).\nPublisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional \naffiliations.\nAuthors and Affiliations\nJingyi Wu1 · Xinting Li2 · Shuikun Li3 · Yanting Luo4 · Ying Gao4 · Yanfei Tao5 · \nJiangying Yang5 · Weiyong Hong6,7\n Weiyong Hong\nweiyongh@126.com\n1 Department of Chinese Medicine (Integrated Chinese and Western Medicine), School of Medicine, Taizhou \nMunicipal Hospital (Taizhou University Affiliated Municipal Hospital), Taizhou University, Taizhou 318000, China\n2 College of Pharmaceutical Science, Zhejiang University of Technology, Hangzhou 310032, China\n3 Department of Pharmacy, Huzhou Maternity & Child Health Care Hospital, Huzhou 313000, China\n4 Zhejiang Moda Biotech Co., Ltd, Hangzhou 310018, Zhejiang, China\n5 Department of Gynaecology and Obstetrics, School of Medicine, Taizhou Municipal Hospital (Taizhou University \nAffiliated Municipal Hospital), Taizhou University, Taizhou 318000, China\n6 Department of Pharmacy, School of Medicine, Taizhou Municipal Hospital (Taizhou University Affiliated \nMunicipal Hospital), Taizhou University, #581 East Shifu Avenue, Taizhou 318000, China\n7 Taizhou Key Laboratory of Infection and Tumor Immunology, Taizhou 318000, China","source_license":"CC0","license_restricted":false}