Exploring the mechanism of Platycladi Cacumen in intervening androgenetic alopecia based on network pharmacology, molecular docking, and molecular dynamics simulation

Abstract

As a traditional hair-growth-promoting herb, Platycladi Cacumen (PC) has a long history of folk application in the field of hair loss improvement. Preliminary modern pharmacological studies have suggested that its active components may exert potential effects by regulating hair follicle-related signaling pathways; however, for androgenetic alopecia (AGA), the exact targets and specific regulatory mechanisms of PC remain unelucidated, which provides a direction for research on natural drug-based intervention in AGA. In this study, network pharmacology was employed to predict the active components and core targets of PC. Targets associated with AGA were collected, and the intersection targets between PC and AGA were identified. Subsequently, protein-protein interaction (PPI) analysis, Gene Ontology (GO) enrichment analysis, and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis were performed on the intersection targets to screen out the core targets. Thereafter, molecular docking and molecular dynamics(MD) simulation were conducted to validate the interactions between key active components and core targets. The component-target network diagram included 1044 interaction relationships between 32 components and 439 targets, among which quercetin(160), apigenin(150), myricetin(129), and hinokinin(105) were identified as key components. The disease-target network diagram summarized 410 targets associated with AGA. Through PPI network analysis, key targets such as ESR1(46), BCL2(44), INS(44), AR(42), and STAT3(40) were screened out. The results of GO enrichment analysis and KEGG pathway analysis revealed that PC may exert its effects by regulating the EGFR receptor molecule and pathways including the HIF-1 signaling pathway. Molecular docking results showed that the binding energies of all complexes were less than -5.0 kcal/mol, indicating favorable binding effects. MD simulation results showed that the root-mean-square deviation (RMSD), root-mean-square fluctuation (RMSF), radius of gyration (Rg), solvent-accessible surface area (SASA), two-dimensional free energy landscape (FEL-2D), and FEL-3D of the simulation system all remained in an equilibrium state with small fluctuation amplitudes. This result indicated that the molecular system had a stable overall conformation, restricted local residue movement, a compact spatial structure, and stable internal chemical bonds–collectively predicting that the quercetin-STAT3, apigenin-AR, myricetin-STAT3, and hinokinin-AR complexes may exhibit satisfactory binding stability. Collectively, Overall, this study systematically investigated the mechanism of action and potential value of PC leaves in intervening in AGA, providing a solid theoretical basis for the intervention of AGA with PC.

Introduction

AGA, the most common type of alopecia in clinical practice, is characterized by miniaturization of the hair follicle, thinning of the hair shaft, and progressive reduction in hair density. Its prevalence varies by region, ethnicity and age: among European males aged 40-49 years¹, the prevalence is 53%, with a lifetime prevalence of up to 90%; the prevalence is relatively lower in Asia², specifically 20% among Chinese males aged 40-49 years and 12% among Chinese females over 70 years³; 14.1% among Korean males and 5.6% among Korean females⁴; and 63% among Singaporean males aged 17-86 years⁵.

AGA is a multifactorial disorder characterized by progressive miniaturization of hair follicles, involving androgen signaling, follicular cell apoptosis, and dysregulation of multiple signaling pathways. Testosterone is converted into dihydrotestosterone (DHT) by 5-reductase^6,7, which activates the AR signaling pathway, a key mechanism in AGA pathogenesis. Additionally, aberrant regulation of Wnt/-catenin, PI3K-Akt, MAPK, and TGF- pathways^8,9, along with inflammation and oxidative stress, collectively contribute to hair follicle cycle disruption. Considering the multi-target nature of AGA, recent studies have increasingly employed in-silico approaches such as network pharmacology, molecular docking, and MD simulations to systematically elucidate its molecular mechanisms and identify potential therapeutic strategies¹⁰ Network pharmacology constructs compound-target-pathway networks¹¹, molecular docking evaluates the binding affinity of active compounds to key targets¹², and MD simulations further assess the stability of these complexes under dynamic conditions¹³. The integration of multiple in-silico methods has become an important paradigm for investigating the effects of natural products on complex diseases such as AGA¹⁴.

Natural products, due to their multi-component and multi-target synergistic effects, have shown potential in alopecia treatment. PC, a traditional Chinese medicinal herb with a long history of use, contains abundant flavonoids, terpenoids, and other bioactive compounds¹⁵. However, the key active constituents of PC and their molecular mechanisms against AGA remain poorly defined, particularly regarding the specificity and stability of compound-target interactions.

In this study, we combined network pharmacology¹⁶, molecular docking, and MD simulations^17,18 to systematically identify the potential active components, key targets, and underlying molecular mechanisms of PC in AGA intervention. This work aims to provide a theoretical basis for developing PC-based natural therapeutics against AGA and to support the modernization of traditional hair-growth medicinal research.

Results

Network pharmacology results of PC

Relevant components of PC were summarized using the Traditional Chinese Medicine Systems Pharmacology Database and Analysis Platform (TCMSP)¹⁹, Herb database²⁰, and Modern Analytical Technology of Chinese Medicine Quality A Reference Manual of the Chinese Pharmacopoeia, followed by component screening in accordance with the screening criteria set in Section Network pharmacology, 32 components of PC were identified (table 1), and these components exhibited 1044 interaction relationships with 439 targets. The above interactions were visualized using Cytoscape 3.10.3 software, generating a component-target network diagram (Fig. 1a); in this diagram, red circles represent targets and orange diamonds represent components. Additionally, the network diagram is ranked by degree value: the larger the degree value, the larger the node, indicating a more critical role of the node in the network diagram. The top-ranked key components identified thereby include quercetin (MOL000098, degree value 160), apigenin (MOL000008, degree value 150), myricetin (MOL002008, degree value 129), and hinokinin (MOL002005, degree value 105)(Fig. 2).

Table 1.

Components of PC.

MOL_ID	molecule_name	mw(Da)	alogp	FASA(Å2)
MOL001055	5-isopropyl-2-methylbicyclo[3.1.0]hex-2-ene	136.26	2.873	0.29
MOL001114	2-[(1R)-2,2,3-trimethyl-1-cyclopent-3-enyl]ethanal	152.26	2.193	0.29
MOL001242	dl-Thujone	152.26	1.771	0.30
MOL001300	2-PHENYLETHANOL	122.18	1.547	0.37
MOL000197	Myrcene	136.26	3.688	0.37
MOL000199	Safrol	162.2	2.606	0.38
MOL000200	(S)-(+)-alpha-Phellandrene	136.26	3.254	0.29
MOL002002	cis-Carveol	152.26	2.401	0.27
MOL002003	(-)-Caryophyllene oxide	220.39	3.519	0.28
MOL002005	Hinokinin	354.38	3.571	0.35
MOL002008	myricetin	318.25	1.236	0.40
MOL002010	Thujopsadiene	202.37	3.687	0.29
MOL002012	2-Phenylpropene	118.19	2.828	0.45
MOL002015	(E,2S)-4-[(1R)-2,6,6-trimethyl-1-cyclohex-2-enyl]but-3-en-2-ol	194.35	3.203	0.27
MOL000202	Moslene	136.26	3.449	0.27
MOL002021	4-Vinyl-m-xylene	132.22	3.354	0.43
MOL002022	(4E,6Z)-2,6-dimethylocta-2,4,6-triene	136.26	3.58	0.36
MOL002028	(+)-beta-Phellandrene	136.26	3.308	0.30
MOL002040	(1S,4R)-fenchone	152.26	2.357	0.29
MOL002042	thymol	150.24	3.243	0.33
MOL000264	Tereben	136.26	3.643	0.28
MOL000268	(1S,5S)-1-isopropyl-4-methylenebicyclo[3.1.0]hexane	136.26	2.927	0.31
MOL000669	(S)-camphor	152.26	1.936	0.27
MOL000708	Benzadehyde	106.13	1.589	0.44
MOL000714	Hyacinthin	120.16	1.517	0.43
MOL000741	(2S,3S)-3,5,7-trihydroxy-2-(4-hydroxyphenyl)chroman-4-one	288.27	1.752	0.41
MOL000008	apigenin	270.25	2.334	0.41
MOL000874	paeonol	166.19	1.286	0.32
MOL000905	()-beta-Pinene	136.26	2.927	0.28
MOL000911	Terpilene	136.26	3.449	0.27
MOL000920	LINALOOL (D)	154.28	2.735	0.32
MOL000098	quercetin	302.25	1.504	0.38

Fig. 1.

Network pharmacology results of PC (a) Component-target network diagram. In this network diagram, the nodes around the periphery represent targets, while the nodes in the center represent. (b) AGA target network diagram. (c) Overlapping target Venn diagram. (d) PPI network diagram.

Fig. 2.

Chemical structures of key compounds.

A total of 410 AGA-related targets were summarized from the GeneCard database²¹ (Fig. 1b). A Venn analysis tool was used to generate an intersection diagram of PC targets and AGA targets (Fig. 1c), yielding 31 overlapping targets. These 31 overlapping targets were imported into the STRING database²² for protein-protein interaction (PPI) network analysis, with the screening criterion set as an interaction score . Subsequently, the obtained data were imported into Cytoscape 3.10.3 software for the construction and optimization of the PPI network diagram. Through analysis using the CytoNCA plugin in Cytoscape, targets with degree centrality (DC) were selected and ranked, ultimately identifying 14 key targets including ESR1, BCL2, INS, AR, STAT3, HIF1A, EGF, and PPARG (Fig. 1d).

GO enrichment analysis²³ (Fig. 3a) showed that the intersecting targets were mainly enriched in functional modules relevant to AGA pathology (threshold: P < 0.05). At the BP level, the enriched terms were primarily related to steroid/androgen-associated metabolic processes and development- and growth-related processes. At the CC level, the enriched terms were mainly associated with nuclear regulatory components, such as the transcription regulator complex. At the MF level, significant enrichment was observed for terms including “steroid hormone receptor binding” and “transcription factor binding/activity”. Collectively, these results suggest that PC-related targets may be closely associated with hormone receptor–mediated transcriptional regulation and hair follicle–related cell fate processes.

Fig. 3.

Enrichment analysis network diagram (a) GO enrichment analysis network diagram. (b) KEGG enrichment analysis network diagram.

KEGG enrichment analysis²⁴ (Fig. 3b) indicated that the intersecting targets were involved in multiple significantly enriched pathways (P < 0.05). Among them,

we prioritized two pathways most directly related to AGA pathogenesis–the MAPK signaling pathway and the HIF-1 signaling pathway–which served as the core framework for subsequent mechanistic integration in the Discussion. In addition, homeostasis-related pathways, including FoxO signaling and autophagy, were also significantly enriched, providing complementary clues for stress- and metabolic-adaptation–related mechanisms.

Molecular docking results

Based on the aforementioned network pharmacology analysis results, this study selected key components (quercetin, apigenin, myricetin, hinokinin) and key targets (ESR1, BCL2, INS, AR, STAT3, HIF1A, EGF, PPARG) for molecular docking experiments. In the generated binding energy heatmap (Fig. 4), a darker color indicates stronger binding stability between the ligands and receptors. The results showed that the docking binding energies between all components and targets were lower than -6.4 kcal/mol, demonstrating that the key components of PC have favorable binding abilities with the key targets of AGA, with ideal docking effects. For each key component, this study selected the docking result with the optimal binding energy to each target for visual display.

Fig. 4.

Binding energy heatmap.

Quercetin–STAT3 (Fig. 5a) and myricetin–STAT3 (Fig. 5c) exhibited similar binding features: the ligands occupy a hydrophobic pocket of STAT3, where hydrophobic anchoring interactions involve residues such as LEU438 and VAL490, complemented by multiple hydrogen bonds with key residues including LYS370, SER381, and THR440, collectively stabilizing the ligand orientation within the pocket. Apigenin–AR (Fig. 5b) displayed a cooperative interaction pattern, combining hydrophobic contacts (e.g., PRO682, LEU744, and LYS808) with hydrogen bonding to GLU681/GLY683, together with – stacking with TRP718 and –cation interactions involving ARG752/LYS808. In contrast, hinokinin–AR (Fig. 5d) was dominated by multiple hydrophobic contacts (e.g., PHE673, LEU674, and ILE841), with an additional backbone hydrogen bond contributing to pose stabilization. To validate the reliability of our docking system, minoxidil, a first-line therapeutic drug for androgenetic alopecia (AGA), was selected as a positive control for molecular docking with AR and STAT3. As shown in Supplementary Fig.3–4, minoxidil exhibited a binding energy of -7.2 kcal/mol with AR and -6.2 kcal/mol with STAT3, which is consistent with previously reported binding affinities of minoxidil to these targets. Notably, the ke -7.8 kcal/mol) with AR and STAT3, which may suggest that these herbal components potentially possess superior binding potential to the core AGA targets compared to the clinical drug minoxidil.

Fig. 5.

Molecular docking visualization and interaction analysis (a) quercetin-STAT3, -9.3 kcal/mol, 8 hydrogen bonds. (b) apigenin-AR, -10.1 kcal/mol, 2 hydrogen bonds. (c) myricetin-STAT3, -8.8 kcal/mol, 7 hydrogen bonds. (d) hinokinin-AR, -7.8 kcal/mol, 1 hydrogen bond.

Overall, these docking models highlight key interaction modes–hydrophobic anchoring, hydrogen-bond networks, and -related interactions–within the STAT3 and AR binding pockets, providing a structural rationale that supports subsequent MD-based stability assessment and mechanistic discussion.

MD simulation results

A 100 ns MD simulation was performed on each of the above four molecular docking complexes to verify their long-term stability. The MD simulation results of quercetin-STAT3 showed that the RMSD of the small molecule was consistently maintained below 0.1 nm, indicating a highly stable conformation. After a brief initial conformational adjustment, the RMSD of the protein remained between 0.3 and 0.4 nm, and the system reached dynamic equilibrium (Fig. 6a). The residue RMSF results revealed only 4 protein residues with an RMSF peak > 0.6 nm, indicating fewer flexible regions and strong overall rigidity (Fig. 6b). The Rg fluctuated stably in the range of 3.40–3.50 nm, and the SASA was maintained at 270–290 nm² without a continuous

Fig. 6.

MD Simulation results of quercetin-STAT3.

changing trend, confirming that both the structural compactness of the complex and its interaction with the solvent remained stable (Fig. 6c, d). A single and concentrated low-energy basin was observed in the free energy landscape (Fig. 6e, f), further verifying the stable energy state of the system.

The MD simulation results of apigenin-AR showed that the RMSD of the small molecule fluctuated around 0.05 nm consistently, indicating a highly stable conformation. After a brief initial conformational adjustment, the RMSD of the protein remained between 0.20 and 0.25 nm, and the system reached dynamic equilibrium (Fig. 7a). The residue RMSF results revealed that most protein residue fluctuations were in a rigid state, demonstrating the stability and reliability of the docking system (Fig. 7b). The Rg fluctuated stably in the range of 1.84–1.88 nm, and the SASA was maintained at 122–136 nm² without a continuous changing trend, confirming that both the structural compactness of the complex and its interaction with the solvent remained stable (Fig. 7c, d). A single and concentrated low-energy basin was observed in the free energy landscape (Fig. 7e, f), further verifying the stable energy state of the system.

Fig. 7.

MD Simulation results of apigenin AR.

The MD simulation results of myricetin-STAT3 showed that the RMSD of the small molecule was consistently maintained below 0.1 nm, indicating a highly stable conformation. After a brief initial conformational adjustment, the RMSD of the protein remained between 0.3 and 0.5 nm, and the system reached dynamic equilibrium (Fig. 8a). The residue RMSF results revealed that most protein residue fluctuations were in a rigid state, demonstrating the stability and reliability of the docking system (Fig. 8b). The Rg fluctuated stably in the range of 3.38–3.48 nm, and the SASA was maintained at 270–295 nm² without a continuous changing trend, confirming that both the structural compactness of the complex and its interaction with the solvent remained stable (Fig. 8c, d). A single and concentrated low-energy basin was observed in the free energy landscape (Fig. 8e, f), further verifying the stable energy state of the system.

Fig. 8.

MD Simulation results of myricetin-STAT3.

The MD simulation results of hinokinin-AR showed that after a brief initial conformational adjustment of the complex’s RMSD, the small molecule reached equilibrium at 40 ns, while the protein remained between 0.20 and 0.30 nm after 20 ns, and the system achieved dynamic equilibrium (Fig. 9a). The residue RMSF results revealed that there were few protein residues with an RMSF peak > 0.6 nm, accompanied by fewer flexible regions and strong overall rigidity, demonstrating the stability and reliability of the docking system (Fig. 9b). The Rg fluctuated stably in the range of 1.83–1.86 nm, and the SASA was maintained at 122–137 nm² without a continuous changing trend, confirming that both the structural compactness of the complex and its interaction with the solvent remained stable (Fig. 9c, d). A single and concentrated low-energy basin was observed in the free energy landscape (Fig. 9e, f), further verifying the stable energy state of the system.

Fig. 9.

MD Simulation results of hinokinin-AR.

These results indicated that all systems achieved dynamic equilibrium within the simulation period, with no severe structural deformation or abnormal energy accumulation observed. There were no sharp peaks, energy divergence, or serious simulation defects such as atomic overlap or chemical bond breakage, demonstrating that the calculation results possess reliable physical significance. To support within-trajectory convergence, we report plateau-region (50–100 ns) statistics and block-averaged stability metrics in Table S1,S1a-S1d.

Discussion

In this study, we propose a systematic strategy integrating network pharmacology, molecular docking^25,26, and MD simulations²⁷ to identify putative active components of PC for intervening AGA and to formulate testable mechanistic hypotheses. It should be emphasized that our conclusions are primarily derived from enrichment analyses and structural simulations; therefore, we favor interpretations framed as “potential involvement/associative evidence” rather than direct inferences of changes in expression or phosphorylation. Overall, the enrichment results suggest that PC-related targets may involve EGFR-associated proliferative signaling²⁸, MAPK pathway branches (ERK/p38)^29–32, and HIF-1–related programs³³, which are linked to dermal papilla cell survival, hair-cycle progression, and follicular microenvironmental adaptation.

This study further expanded the target-level interpretation of two hub targets, AR and STAT3. AR is the core receptor of the canonical DHT–AR pathogenic axis and is closely associated with follicular miniaturization and hair-cycle dysregulation, mainly through androgen-responsive downstream transcriptional programs³⁴. Accordingly, considering the hub-like role of AR in our network and the enrichment of steroid hormone receptor–binding terms, we infer that PC may modulate AGA-relevant androgen signaling at the receptor/transcriptional regulatory level. In parallel, STAT3, as an inflammation- and stress-associated transcription factor, has been implicated in hair follicle homeostasis and hair-cycle regulation, and is linked to survival and regenerative responses of follicular epithelial stem/progenitor cells^35,36. Given that STAT3 can integrate inflammatory/stress cues and crosstalk with signaling such as MAPK, its hub status provides a biologically interpretable node connecting the MAPK and HIF-1 enrichments observed in this study with inflammation/stress-related changes in the follicular microenvironment in AGA.

Taken together, we propose a more focused and testable hypothesis: core bioactive components of PC may modulate key nodes such as AR and STAT3, thereby coordinately influencing MAPK- and HIF-1–linked transcriptional responses, participating in the regulation of hair-cycle progression, maintenance of follicular stem/progenitor cell function, and modulation of inflammation/stress responses, and ultimately affecting AGA-related phenotypes. Finally, molecular docking and MD simulations provide structural support for the feasibility and stability of interactions between key active components and core targets, offering a rationale for prioritizing compound–target pairs for subsequent experimental validation.

Materials and methods

Software and databases

Software: pymol(version 2.6.0); autodock tools (version 1.5.6); gromacs (version 2023.3); sobtop (version 1.0); OpenBabel (version 3.1.1); SPDBV (version 4.10); sobtop (version 1.0); Cytoscape (version 3.10.3); Visual Studio Code (version 1.102.1).

Databases: TCMSP https://www.tcmsp-e.com/(accessed on 8 July 2025); Herb http://herb.ac.cn/(accessed on 9 July 2025); SwissTargetPredicition https://www.swisstargetprediction.ch/(accessed on 10 July 2025); Gene Card https://www.genecards.org/(accessed on 14 July 2025); Pubchem https://pubchem.ncbi.nlm.nih.gov/(accessed on 15 July 2025); Uniprot https://www.uniprot.org/(accessed on 15 July 2025); RCSB PDB https://www.rcsb.org/(accessed on 18 July 2025); String https://cn.string-db.org/(accessed on 25 July 2025); Metascape https://metascape.org/(accessed on 30 July 2025); KEGG https://www.kegg.jp/(accessed on 26 July 2025); bioinformatics https://bioinformatics.com.cn/(accessed on 30 July 2025).

Experimental procedures

Network pharmacology

Combined with the literature and Modern Analytical Technology of Chinese Medicine Quality A Reference Manual of the Chinese Pharmacopoeia, the TCMSP was used to screen the components of PC with the following criteria: calculated lipophilic-water partition coefficient (alogP) between 1.0 and 4.0, molecular weight (mw) between 100 Da and 500 Da, and accessible surface area (FASA) between 0.27 and 0.50^37–39. Among these criteria, the alogP value balances lipophilicity and hydrophilicity to ensure transdermal penetration; the mw complies with the “500 Dalton Rule” for stratum corneum diffusion and guarantees local retention. By integrating the screening results from the TCMSP and the SwissTargetPrediction database, we finally obtained the set of potential targets corresponding to the active components of PC. Specifically, all associated targets were directly included from the TCMSP database without additional screening; whereas only the target genes with non-zero probability scores were retained from the SwissTargetPrediction database, which effectively reduces the risk of false positives. The targets of PC were identified using the Uniprot and TCMSP databases, while the gene names corresponding to AGA were compiled via the Genecard database, with only those targets with non-zero Relevance scores retained to enhance data reliability. Cytoscape 3.10.3 software was used to construct network diagrams, including component-target and disease-target networks. After network topological analysis, the diagrams were plotted based on the degree value: a larger degree value corresponded to a larger node size, indicating a more important role of the node. This approach was used to screen the main components of PC. The overlapping targets between PC and AGA were identified using the MicroBiotech Platform. The PPI network diagram of these overlapping targets was constructed in the STRING database, with the species restricted to Homo sapiens and a confidence score threshold of 0.4. The potential PPI network of target proteins was then generated and sorted by DC: a darker color and larger node size indicated a more critical role of the target. This method was applied to screen the key targets. The overlapping targets between PC and AGA were imported into the Metascape database for GO biological process analysis and KEGG enrichment analysis, with a significance threshold of P < 0.05. GO analysis covered three dimensions: BP, CC, and MF; the top 10 terms in each dimension were selected by ascending order of P-values. For KEGG enrichment analysis, pathways were sorted by ascending order of P-values, and the top 20 pathways were selected for further analysis.

Molecular docking

Small-molecule ligands were downloaded from the PubChem database, and relevant protein receptors were obtained from the RCSB PDB database. The three-dimensional structure of the biomacromolecular receptor was preprocessed using PyMOL software to remove crystal water and free atoms from the structure. The structure of small-molecule ligands was optimized via ChemOffice software to ensure that the ligands and receptors were in a suitable initial state for docking. Subsequently, molecular docking was performed using AutoDock Vina software: this software can simulate the binding process of ligands in the active pocket of the receptor, calculate the binding energy under different binding modes, and screen out the dominant conformations with strong binding affinity based on the binding energy. Finally, the specific binding sites of molecular docking were identified by combining PyMOL software with the Protein-Ligand Interaction Profiler. STAT3 (PDB ID: 6qhd⁴⁰) exhibits a unique post-translational modification state–lysine acetylation and tyrosine phosphorylation–along with the corresponding active binding conformation. This feature precisely matches the regulatory mechanism of STAT3 activity in AGA, providing a molecular model that more closely reflects the physiological and pathological state for the investigation of traditional Chinese medicine (TCM) active component interventions. In this study, Stattic, a classic inhibitor of 6qhd, was selected as the validation ligand for molecular docking. The validation results showed a heavy-atom RMSD of 0.014 Å, confirming the reliability of the docking method. AR (PDB ID: 1t7t⁴¹) accurately recapitulates the core molecular mechanism underlying AGA pathogenesis–the binding of DHT to AR and subsequent activation of downstream signaling pathways–thus serving as the most direct target model for TCM active component interventions in AGA. In this study, the native co-crystal ligand DHT was extracted from the 1t7t structure, preprocessed using AutoDock Tools, and re-docked with the same parameters as those used for subsequent TCM active component screening. The resulting RMSD value was 0.29 Å, further validating the rationality of the entire docking system⁴² (Supplementary Fig.1-2).

Molecular dynamics simulation

To evaluate the binding stability from molecular docking, MD simulations were performed on the protein-small molecule complexes obtained from molecular docking using GROMACS 2023.3 with the AMBER99SB biomolecular simulation package in the Linux system. The small-molecule ligands were parameterized using Sobtop⁴³, and the GAFF (General AMBER Force Field) was adopted to generate ligand topologies and force-field parameters; partial atomic charges were assigned using the AM1-BCC scheme^44,45 . The ligand parameters were then converted into formats compatible with the GROMACS/AMBER force-field framework and combined with AMBER99SB and the SPC explicit water model to assemble the complete simulation systems. Finally, the ligand topology and coordinate files were used for subsequent energy minimization, equilibration, and production MD simulations. For the STAT3 protein, the triclinic unit cell function was adopted, and the dimensions of the simulation box were set along the x, y, and z axes. The SPC (Single Point Charge) water model was used to create an aqueous solution environment. Na+ and Cl- were added to the simulation system to make the system electrically neutral^46,47. A loose convergence threshold of 1000 kJ/(molnm) was initially employed for preliminary energy minimization to relax severe steric hindrances within the system. Subsequently, a rigorous two-step energy minimization strategy was implemented for further optimization: the steepest descent algorithm was first applied to converge the maximum force to 100 kJ/(molnm), followed by the conjugate gradient algorithm to achieve final convergence with a threshold of 50 kJ/(molnm)⁴⁸. This stepwise minimization protocol effectively fully relaxes local steric conflicts in the protein backbone, side chains, and ligand-binding pocket of the complex, while avoiding structural distortion caused by excessive force in a single minimization step. After energy minimization, the system temperature was gradually heated to 300 K. Position restraints were then applied with a force constant of 1000 kJ/(molnm) along the X, Y, and Z axes, and the minimum distance between the solute and the box boundary was set to 1.0 nm. Subsequently, 1000 ps of NVT equilibration and 2000 ps of NPT equilibration were performed at 300 K. Finally, under periodic boundary conditions, unrestrained MD simulation was conducted for 100 ns at constant temperature and pressure, with trajectory data saved every 2.0 ps. The RMSD, RMSF, Rg, SASA, FEL-2D, and FEL-3D landscape of MD simulations were calculated using Python code with Visual Studio Code 1.102.1 software.

Conclusions

In this study, we integrated network pharmacology, molecular docking, and MD simulation techniques to systematically dissect the potential active components and core regulatory network of PC in the intervention of AGA. Our results identified AR and STAT3 as key hub targets for PC-mediated AGA intervention. GO/KEGG enrichment analyses revealed that candidate targets were centrally associated with core biological processes including hormone receptor-mediated transcriptional regulation, hair follicle cell fate determination, and hair follicle microenvironmental homeostasis, and were significantly enriched in the DHT-AR axis, MAPK and HIF-1 signaling pathways. Molecular docking and MD simulations further indicated that the core active components of PC can form stable, high-affinity complexes with AR and STAT3, and provided structural validation for the feasibility of key “compound-target” interactions. This study is the first to construct a “active component-hub target-key pathway” regulatory framework for PC in AGA intervention, breaking the limitations of traditional single-component/single-target research. It provides a new computational biology paradigm for investigating the mechanisms of natural products against AGA and offers precise insights for the development of novel targeted candidate drugs. Given that the conclusions of this study are based on computational predictions and hypothesis generation, future research will require benchmarking against clinically used AGA drugs, in vitro hair follicle cell experiments, and in vivo animal model validation to clarify the biological effects and therapeutic potential of PC, thereby facilitating its clinical translation.

Limitations and future directions

In future work, we will (i) perform benchmark recovery analyses against clinically used AGA drugs and well-established targets/pathways, (ii) leverage public transcriptomic/proteomic datasets to assess expression consistency of hub targets in AGA, and (iii) conduct in vitro and in vivo experiments (e.g., dermal papilla cell assays and relevant models) to validate the predicted compound–target mechanisms and their effects on AGA-related phenotypes.

Author contributions

J.L.: Conceptualization, Methodology, Software, Validation, Data Curation, Writing–Original Draft, Visualization. H.M.: Methodology, Formal Analysis, Investigation, Writing–Review and Editing, Supervision, Project Administration. C.R.: Methodology, Investigation, Resources, Writing–Review and Editing, Supervision, Project Administration. Y.M.: Conceptualization, Methodology, Software, Formal Analysis, Investigation, Resources, Data Curation, Supervision. All authors have reviewed the manuscript.

Funding

Not applicable.

Data availability

The chemical, target, and biological data used in this study were retrieved from the following public databases: Traditional Chinese Medicine Systems Pharmacology Database and Analysis Platform (TCMSP, https://www.tcmsp-e.com/, accessed on 8 July 2025); Herb Database (http://herb.ac.cn/, accessed on 9 July 2025); SwissTargetPrediction (https://www.swisstargetprediction.ch/, accessed on 10 July 2025); GeneCards (https://www.genecards.org/, accessed on 14 July 2025); PubChem (https://pubchem.ncbi.nlm.nih.gov/, accessed on 15 July 2025); UniProt (https://www.uniprot.org/, accessed on 15 July 2025); RCSB Protein Data Bank (RCSB PDB, https://www.rcsb.org/, accessed on 18 July 2025); STRING (https://cn.string-db.org/, accessed on 25 July 2025); Metascape (https://metascape.org/, accessed on 30 July 2025); Kyoto Encyclopedia of Genes and Genomes (KEGG, https://www.kegg.jp/, accessed on 26 July 2025); and Bioinformatics (https://bioinformatics.com.cn/, accessed on 30 July 2025). All data supporting the findings of this study are available within the manuscript and its Supplementary Information.

Code availability

Custom codes used in this study are available from the corresponding author upon reasonable request. No restrictions apply to the use of these codes for non-commercial research purposes.

Declarations

Competing interests

The authors declare no competing interests.

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-37638-0.