Abstract

Background:

Idiopathic pulmonary fibrosis (IPF) is a progressive and fatal lung disease with limited therapeutic options. This study aims to identify potential efficacy biomarkers of Jingfang Granule (JFG) and investigate the therapeutic mechanism of its key active components against critical targets in IPF.

Methods:

IPF-related targets were identified through bioinformatics analysis of a public IPF dataset. Co-expressed gene modules were identified using Weighted Gene Co-expression Network Analysis (WGCNA). The JFG-PF interaction network was constructed employing protein-protein interaction (PPI) methods and functional enrichment analysis, which included Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG), and Gene Set Variation Analysis (GSVA). Additionally, five types of machine learning techniques were utilized to obtain diagnostic markers, which were further analyzed in immune infiltration assessments to evaluate their associations with immune cells and potential therapeutic effects. Molecular docking and molecular dynamics simulations were conducted to validate these analytical results. Finally, immunohistochemistry and immunofluorescence were used for verification.

Results:

A total of 3, 360 upregulated and 240 downregulated differentially expressed genes (DEGs) were identified in IPF sample. Integration of WGCNA findings with 1, 077 targets of JFG pinpointed 57 key genes lingking with IPF and JFG. Machine learning algorithms further refined this list, identifying four diagnostic markers:SPP1, MMP1, AKR1B10, and HTR2A. Immune infiltration analysis revealed that these biomarkers are significantly correlated with alterations in multiple immune cell populations within the IPF microenvironment. Molecular docking experiments strong binding affinities between the active compounds of JFG and these protein biomarkers. Subsequent in vitro validation results demonstrated that JFG significantly regulates the expression of SPP1 and MMP1 in lung fibroblasts, thereby attenuating fibrotic responses through the suppression of pro-fibrotic pathways.

Conclusion:

The study identified four key genes as potential diagnostic markers for IPF and therapeutic targets for JFG. The finding preliminarily elucidate ther mechanisms of JFG in mitigating pulmonary fibrosis vis regulation of fibrotic pathways and the immune microenvironment, thereby providing an integrativeevidence chain for the development of novel-fibrotic therapeutic.

1 Introduction

Idiopathic Pulmonary Fibrosis (IPF) is a progressive and fatal interstitial lung disease characterized by irreversible pulmonary fibrosis, with a median survival of only 2–5 years post-diagnosis (Honda et al., 2023; Fujimoto et al., 2015). With the intensification of population aging, the global burden of this disease continues to increase. Its core pathogenesis is attributed to abnormal repair responses triggered by repeated micro-injuries to the alveolar epithelium in genetically susceptible individuals (Jiang et al., 2025; Selman et al., 2019; MacNee et al., 2014). In this process, the dysregulation of repair mechanisms leads to the persistent activation of fibroblasts and myofibroblasts (Jiang et al., 2025). Resulting in excessive deposition of extracellular matrix and the formation of fibrous scars, ultimately leading to progressive destruction of lung tissue structure (Pan et al., 2025; Fijardo et al., 2024). Recent studies have revealed that immune dysregulation plays a crucial role in the pathogenesis of IPF, with aberrant activation of both the innate and adaptive immune systems involved in the entire process of epithelial injury, fibroblast activation, and ECM remodeling. The complex interactions between immune cells and structural cells create a self-perpetuating fibrotic microenvironment. Current treatment strategies for IPF include anti-fibrotic drugs such as nintedanib and pirfenidone, which can partially slow the decline in lung function (Gao et al., 2024). However, these treatments still face challenges, including (Glass et al., 2022; Chianese et al., 2024).

Currently, Traditional Chinese Medicine (TCM) is widely accepted in clinical practice due to its characteristics of multiple components, multiple targets, and minimal side effects, particularly demonstraes unique advantages in regulating the immune system and treating chronic inflammatory diseases (Abebaw et al., 2025; Reddy and Fan, 2022). Jingfang Granule (JFG) is a modern preparation derived from Jingfang Baidu San, as recorded in the Ming Dynasty herbal formulary “Shesheng Zhongmiao Fang”. This medicine has been included in the Class B of the National Reimbursement Drug List (2021 Edition). It is consists of eleven herbs:Jingjie (Schizonepeta tenuifolia (Benth.)Briq), Fangfeng (Saposhnikovia divaricate (Turcz.)Schischk), Qianghuo (Notopterygium incisum), Chaihu (Bupleurum chinense DC), Qianhu (Peucedanum praeruptorum Dunn), Chuanxiong (Ligusticum striatum DC), Zhiqiao (Aurantii Fructus L), Fuling (Poria cocos (Schw.)Wolf), Jiegeng (Platycodon grandiflorus (Jacq.)A. DC) and Gancao (Glycyrrhiza uralensis Fisch. Ex DC), JFG is utilized in the onset of new pneumonia and is clinically applied in the treatment of various epidemic and infectious diseases, including acute viral upper respiratory infections, dengue fever, influenza A (H1N1), chickenpox, and mumps (Zhu et al., 2022; Li et al., 2023). Previous studies have demonstrated that JFG exhibits significant therapeutic effects on CCl4-induced liver fibrosis, potentially related to the reduction of pro-inflammatory factor expression, enhanced antioxidant activity, and modulation of the TGF-β/Smad4 signaling pathway (Yue et al., 2025). Additionally, JFG can alleviate bleomycin-induced acute lung injury by inhibiting the NF-κB signaling pathway and promoting the generation of regulatory T cells (Tregs). Preliminary animal experiments suggest that the polysaccharide components in JFG may mitigate fibrosis by suppressing the TGF-β1/Smad3 pathway and reducing collagen deposition (Jiao et al., 2025; Yao et al., 2019; Chen et al., 2024). Furthermore, volatile oil components in JFG, such as perilla oil, may regulate the immune microenvironment and protect alveolar epithelial cells (Igarashi and Miyazaki, 2013; Yi et al., 2025). These findings indicate that JFG has potential therapeutic effects on IPF. However, the specific efficacy of JFG in treating pulmonary fibrosis, as well as its active components, molecular targets, pathway mechanisms, and clinical evidence in patients with IPF, necessitate further in-depth research.

To address the aforemidentionfied gap, this study aims to systematically identify the key therapeutic targets of JFG in anti-fibrosis and elucidate its molecular mechanisms of action. We employed integrated bioinformatics analysis of IPF gene expression profiles to screen for differentially expressed genes and co-expression modules. This analysis was combined with JFG component target identification to uncover shared co-expression gene modules between the drug and the disease. Subsequently, we utilized five machine learning algorithms to screen core diagnostic biomarkers from the intersecting targets. Finally, we conducted further validation through immune infiltration, molecular docking, and molecular dynamics simulations. The findings reveal the key targets and pathways through which JFG alleviates pulmonary fibrosis from multiple dimensions, providing a mechanistic basis for its clinical application.

2 Materials and methods2.1 Data collection

The IPF dataset utilized in this study was acquired from the GEO database (https://www.ncbi.nlm.nih.gov/geo/) using the search keyword’pulmonary fibrosis’. The resulting dataset, GSE10667, comprises lung tissue samples from 23 stable IPF patients, 8 patients with acute exacerbation of IPF, and 15 normal controls from disease-free tissues (Rosas et al., 2008). The second dataset, GSE53845, includes lung tissue samples from 40 IPF patients and 8 healthy controls . These two datasets constitute the basis for the subsequent differential expression analysis and bioinformatics mining in this study.

2.2 Identification of differentially expressed genes

The raw microarray data were normalized and subjected to quality control using the limma package (version 3.4.6) in R (version 4.4.3). Differential expression analysis was conducted based on the criteria of adjusted p < 0.05 and|log2 fold change (FC)| ≥1.2 to identify differentially expressed genes (DEGs) (Love et al., 2014). Significantly differentially expressed genes were visualized through volcano plots and hierarchical clustering heatmaps. Functional enrichment analysis was performed on DEGs using the clusterProfiler package (version 3.14.3) along with the org. Hs. e.g., db annotation (version 3.1.0) to identify significantly enriched Gene Ontology (GO) terms and KEGG pathways (p < 0.05) (Li and Maierheba, 2024). A brief workflow of this study is illustrated in Figure 1.

The complete research workflow.

2.3 Weighted gene co-expression network analysis

To investigate the association between gene expression levels and diseases, we constructed a co-expression network using the “WGCNA” package (version 1.71) (Love et al., 2014). Following hierarchical clustering to eliminate outlier samples, genes exhibiting a median absolute deviation of expression variability ranking in the lowest 50% were excluded from the analysis. Based on the threshold established by the pickSoftThreshold function, the minimum power that achieved a scale-free topology fit index (R2) greater than 0.85 was identified, leading to the construction of an unsigned scale-free network. The adjacency matrix was then converted into a Topological Overlap Matrix (TOM), and gene modules were delineated through dynamic pruning (with a minimum module size of 30). Modules exhibiting eigengene dissimilarity of less than 0.25 (correlation coefficient >0.75) were merged. The relationship between modules and phenotypes was quantified by calculating Gene Significance (GS) and Module Membership (MM). The grey module (comprising unclassified genes) was excluded from all subsequent analyses.

2.4 Acquisition of Jingfang Granule targets

Eleven components of JFG, namely, Schizonepeta, Saposhnikovia, Bupleurum, Peucedanum, and other Chinese medicinal herbs, were retrieved from TCMSP (version 2.3; https://www.tcmsp-e.com/) (Ru et al., 2014). Based on the pharmacokinetic parameters provided by TCMSP, filtering criteria of OB ≥30% and DL ≥0.18 were applied to identify the qualified active ingredients from these herbal components (Xu et al., 2012). Searched for the target proteins corresponding to each active ingredient. To identify eligible active ingredients not included in TCMSP, we also consulted the literature and obtained standard SMILES structures and bioassay-confirmed targets from the PubChem database (https://pubchem.ncbi.nlm.nih.gov) (Maphetu et al., 2022). These SMILES numbers were submitted to the SwissTargetPrediction database (http://www.swisstargetprediction.ch) to predict potential target proteins of bioactive components (p > 0.4). Finally, utilize the ID mapping function of the UniProt database to standardize and deduplicate the target genes, integrate target information from all sources, and ultimately obtain the complete drug target set of JFG.

2.5 Construction of protein-protein interaction (PPI) network and drug–disease networks

The previously described DEGs, IPF-related targets from WGCNA, and predicted drug targets of JFG were intersected using the Venny 2.1 tool (https://bioinfogp.cnb.csic.es/tools/venny/) to identify the core therapeutic targets of JFG for IPF treatment, which were reserved for subsequent functional validation.

The resulting cross-targets were then used to construct a PPI network within the context of JFG-IPF interactions via the STRING database (https://string-db.org/) (Szklarczyk et al., 2019) with the organism limited to Homo sapiens and a combined score threshold of ≥0.7. The entire network, consisting of all interacting proteins that met the confidence threshold, was retained. Finally, the PPI graph was visualized using the network analyzer plugin in Cytoscape 3.9.0, thus constructing the JFG-IPF network map (Shannon et al., 2003; Assenov et al., 2008).

2.6 Functional enrichment analysis

To explore the potential biological functions and major signaling pathways of JFG in treating IPF, we conducted GO (Wu et al., 2021a) functional and KEGG pathway enrichment analyses on the intersection targets using the clusterProfiler package in R software (version 4.4.3) (Dennis et al., 2003; Chang et al., 2025). Both GO and KEGG analyses utilized significance thresholds of p-value<0.05. This analysis framework offers critical insights into the functional roles and molecular pathways of IPF.

2.7 Machine learning to identify target hub genes

To identify hub genes among the intersecting targets, five machine learning algorithms—Least Absolute Shrinkage and Selection Operator (LASSO), Random Forest (RF) (Hu and Szymczak, 2023), Support Vector Machine-Recursive Feature Elimination (SVM-RFE), K-Nearest Neighbors (KNN) (Sanz et al., 2018), and NeuroEvolution of Augmenting Topologies (NEAT)-were employed. Overlapping genes identified by all five algorithms were subsequently defined as core therapeutic targets of JFG for treating pulmonary fibrosis. Further gene correlation analysis and differential expression validation were conducted to elucidate their potential functions and clinical significance.

2.8 External dataset validation

To achieve cross-species analysis and further validate the research findings, this study employed the “homologene” package in R software to perform homologous conversion of hub genes into their corresponding mouse genes. Following the conversion, the expression levels of these hub genes were validated using external datasets. This validation process is crucial for confirming the relevance and importance of the identified hub genes across different biological contexts, thereby enhancing the credibility and robustness of the study results.

2.9 Immune infiltration analysis

To uncover the potential connections between diagnostic markers and immune cells, the R package ‘GSVA’ (version 1.44.0) was employed to estimate immune cell infiltration using the gene signature-based method xCell, This approach explored the relationship between varying levels of immune cell infiltration and the diagnostic markers for IPF. Furthermore, the correlations between hub genes and immune cells were assessed through Spearman correlation.

2.10 Analysis of gene set enrichment analysis for single genes

The present study employed single-gene gene set enrichment analysis (GSEA) to elucidate the functional roles of the screened biomarkers in the IPF conditions and to identify their putative downstream targets. Samples were stratified into high-and low-expression groups based on the median expression levels of the two genes of interest. The “c2.cp.kegg_legacy.v2025.1. Hs. symbols. gmt” gene set database was used as the reference standard.

2.11 Molecular docking

The structural information of the target hub protein was obtained from the PDB database (https://www.rcsb.org/), while the structure of the active ingredient was sourced from the PubChem database (Kim et al., 2021). These structural files underwent preprocessing utilizing AutoDock software (version 4.2.6), which included the removal of water molecules and the addition of hydrogen atoms, followed by conversion into the PDBQT format (Burley, 2025). Subsequently, we analyzed the binding sites of the target hub protein to identify the corresponding docking active pockets. The preprocessed protein and ligand files were imported into AutoDock Vina (version 1.1.2) for docking validation. The output results were then visualized in a heatmap to illustrate the potential binding affinities of these key bioactive compounds with the target hub protein. Finally, several representative docking diagrams were rendered using PyMOL (version 1.7.0; https://pymol.org/).

2.12 Molecular dynamics simulation

Molecular dynamics (MD) simulations were conducted using GROMACS version 2023.4. The CHARMM36 all-atom force field was employed for modeling the protein receptor, while the TIP3P model was utilized for the explicit representation of water molecules (Loubet et al., 2014). The ligand topology and parameters were generated using the CHARMM Generalized Force Field (CGenFF) through the CGenFF online server (https://cgenff.umaryland.edu/). The systems were solvated in a cubic water box, ensuring a minimum distance of 1.2 nm between the solute and the box edge. Counterions (Na+/Cl−) were introduced to neutralize the system’s charge and to establish a physiological salt concentration of 0.15 M (Milla, 2022). The system’s energy was initially minimized using the steepest descent algorithm until the maximum force dropped below 1,000 kJ/mol/nm. Following this, a two-step equilibration process was implemented:first, a 100 ps NVT equilibration at 300 K utilizing the V-rescale thermostat, and second, a 100 ps NPT equilibration at 1 bar employing the Parrinello-Rahman barostat. During both equilibration phases, position restraints were applied to the protein’s heavy atoms (da Fonseca et al., 2024). Finally, the binding free energy of the protein-ligand complex was calculated using the MM/PBSA method to evaluate its binding affinity.

2.13 Animal experiment

The animal protocol used in this study was approved by the Ethics Committee of the Laboratory Animal Center at Heilongjiang University of Traditional Chinese Medicine (Resolution No. 2025050935). Male C57BL/6 mice (8 weeks old, 20–25 g) were obtained from Beijing Viton–Lever Laboratory Animal Technology Co., Ltd. (Beijing, China) (SCXK–(Beijing) 2021–0006) and housed under controlled conditions (20 ± 2 °C, 40–70% humidity, 12–hour light/dark cycle) with free access to food and water. After a one-week acclimatization period, 60 mice were randomly divided into six groups (10 mice per group): control group, BLM group, BLM+JFG (low, medium, and high dose) groups, and BLM+PFD group. For intratracheal instillation, mice were anesthetized via intraperitoneal injection of sodium pentobarbital (90 mg/kg). The depth of anesthesia was confirmed by the absence of a pedal withdrawal reflex. The control group received a single intratracheal instillation of 0.9% saline, while all other groups received an intratracheal injection of bleomycin (BLM, 5 mg/kg) to induce pulmonary fibrosis (Gul et al., 2023). Starting on the day of BLM/saline administration, mice were administered JFG (1.01, 2.01, and 3.03 g/kg/day) or pirfenidone (PFD, 300 mg/kg/day) orally twice daily for 21 consecutive days.

All mice were humanely euthanized within 24 hours of the final dose. Euthanasia was performed via intraperitoneal injection of a lethal dose of sodium pentobarbital (150 mg/kg). Death was confirmed by cardiac arrest and loss of corneal reflex, followed by cardiac perfusion with pre–chilled PBS to clear residual blood from the lungs. The lungs were then excised; the left lung was used for histopathological examination, and the right lung was rapidly frozen at –80 °C for subsequent analysis. All anesthesia and euthanasia procedures were performed in accordance with the AVMA Guidelines for the Euthanasia of Animals (2020 edition).

2.14 Morphological analysis

The isolated lung tissues of mice were fixed in 4% polyformaldehyde for 24 h. After fixation, the tissues were embedded in paraffin and sectioned into 5 μm thick slices. The sections were stained with hematoxylin-eosin (HE) staining kit, and histopathological changes were observed under optical microscopy, with particular attention to inflammatory infiltration, structural alterations, and collagen deposition.

2.15 ELISA measurement of hydroxyproline (HYP)

The collagen content in mouse lung tissues was assessed using a Hydroxyproline Assay Kit (Nanjing Jiancheng Bioengineering Institute, China). Approximately 30 mg of lung tissue was accurately weighed and hydrolyzed according to the manufacturer’s protocol. The hydrolyzed samples were neutralized by adjusting the pH to a range of 6–8 to ensure optimal detection conditions. After centrifugation, the supernatant was collected, and the absorbance was measured at a wavelength of 550 nm using a spectrophotometer. The hydroxyproline content was quantified based on a standard curve to evaluate the degree of collagen deposition in the lung tissues, reflecting the extent of pulmonary fibrosis.

2.16 ELISA measurement of inflammatory factors

The levels of inflammatory factors in mouse lung tissues were assessed using detection kits for IL-6, IL-1β, and IL-10 (Nanjing Jiancheng Bioengineering Institute, China). Approximately 30 mg of lung tissue was accurately weighed. Tissue homogenates were prepared by homogenization and lysis according to the manufacturer’s instructions. The pH of the samples was adjusted to the range of 6.0–8.0 to meet the subsequent detection conditions. After centrifugation, the supernatant was collected, and its absorbance was measured at a wavelength of 550 nm using a spectrophotometer. The concentrations of IL-6, IL-1β, and IL-10 in the samples were calculated using standard curves to evaluate pulmonary inflammation levels, thereby reflecting the progression of pulmonary fibrosis.

2.17 Immunohistochemistry (IHC) and immunofluorescence (IF)

For both IHC and IF analyses, serial 4 μm paraffin-embedded tissue sections were subjected to a standardized procedure. Briefly, sections were baked at 70 °C for 2 h, deparaffinized in xylene, and rehydrated through a graded alcohol series. Antigen retrieval was performed using citrate buffer (pH 6.0). For IHC staining specifically, endogenous peroxidase activity was blocked with 0.3% hydrogen peroxide. All sections were then blocked with 2.5% sheep serum to minimize non-specific binding and subsequently incubated with primary rabbit antibodies against Collagen I (67288-1-Ig, Proteintech, Wuhan, China, 1:5,000), α-SMA (14395-1-AP, Proteintech, Wuhan, China, 1:2000), SPP1 (HA723601, HUABIO, China) and MMP1 (ER31211, HUABIO, China) at 30 °C for 80 min. For IHC detection, the sections were incubated with an HRP-conjugated secondary antibody, visualized with DAB chromogen, counterstained with hematoxylin, and mounted with neutral gum. For parallel IF detection, the sections were instead incubated with an Alexa Fluor 488-conjugated goat anti-rabbit secondary antibody, with nuclei counterstained using DAPI, and finally mounted with an anti-fade mounting medium.

2.18 Cell culture and processing

The human fetal lung fibroblast cell line MRC-5 was obtained from Wuhan Punosai Life Sciences Co. The cells were cultured at 37 °C in a humidified incubator with 5% CO2 using MRC5 cell-specific medium. Cells were allowed to grow to 60%–70% confluence and then serumstarved in a serum-free medium for 6–8 h. Recombinant human TGF- β1 (10 ng/mL, PeproTech, USA) was added for 24 h to stimulate the cellular lung fibrosis model.

2.19 Cell viability

Cell proliferation was assessed using the Cell Counting Kit-8 (CCK-8, Dojindo, Kumamoto, Japan) according to the manufacturer’s instructions. Briefly, MRC-5 cells were inoculated into 96-well plates, stimulated with TGF-β1, and treated with increasing concentrations of JFG. Cells were incubated with CCK-8 dissolved in the medium for 3 h at 37 °C, and the absorbance at 450 nm was measured using an enzyme reader (Thermo MMLTISKAN GO, USA).

2.20 Western blot analysis

After cell lysis, total proteins were extracted using RIPA lysis buffer (C1053, Prilosec, Beijing, China), and protein concentration was determined by BCA protein assay kit (Cat. No. P1511, Prilosec, Beijing, China). According to the molecular weight differences of target proteins, SDS-PAGE electrophoresis separation was conducted using 10% polyacrylamide gel (24 μg protein per lane), followed by transfer onto PVDF membranes. Due to the wide molecular weight range of target proteins (36–139 kDa), each target protein was analyzed on independent membranes to ensure optimal separation and detection sensitivity. After blocking with rapid blocking buffer at room temperature for 20 min, the membranes were incubated overnight at 4 °C with the following primary antibodies: Collagen I (67288-1-Ig, Proteintech, Wuhan, China, 1:5,000), E-cadherin (20874-1-AP, Proteintech, Wuhan, China, 1:5,000), ACSL4 (22401-1-AP, Proteintech, Wuhan, China, 1:5,000), SPP1 (HA723601, HUABIO, China, 1:2000), Elastin (15257-1-AP, Proteintech, Wuhan, China, 1:1,000), Occludin (27260-1-AP, Proteintech, Wuhan, China, 1:1,000), MMP1 (ER31211, HUABIO, China, 1:2000), Vimentin (10366-1-AP, Proteintech, Wuhan, China, 1:5,000), α-SMA (14395-1-AP, Proteintech, Wuhan, China, 1:2,000). To ensure that each target protein had its own specific loading control from the same membrane, the antibodies on each membrane were stripped using an enhanced stripping buffer (antibody stripping solution, enhanced) after target protein detection. The same membranes were then reprobed with GAPDH antibody (60004-1-Ig, Proteintech, Wuhan, China, 1:5,000) and visualized following the same procedure. Band density was quantified using ImageLab software (Bio-Rad), with target protein expression levels normalized to the GAPDH signal obtained from the identical membrane after stripping and reprobing.ormalized to the specific internal reference GAPDH control from the same blot experiment.

2.21 Statistical analyses

All data processing and analysis were conducted using R software (version 4.4.3) unless otherwise stated. p < 0.05 was considered statistically significant, and the significance level is denoted as follows: *(p < 0.05), **(p < 0.01), and ***(p < 0.001) by one-way ANOVA.

3 Results3.1 3600 shared dysregulated genes were identified in IPF

Identifying 3,360 upregulated and 240 downregulated DEGs (Supplementary Table S1). The distribution of DEGs was visualized through a volcano plot (Figure 2A) and a hierarchical clustering heatmap (Figure 2B). Subsequently, a gene correlation network was constructed among the key DEGs (Figure 2C), further illustrating the distinct expression patterns of differential genes between the disease group and the control group. Network topology analysis revealed strong correlations among the pathways enriched by these differential genes (Figure 2D). All data were sourced from the publicly available GEO database, thus exempting this study from additional ethical review.

Acquisition of IPF DEGs in the GSE10667 dataset. (A) Heatmap of DEGs analysis results. (B) Volcano plot of DEGs analysis results. (C) Correlation among the top 20 DEGs. (D) Network topology of the top 20 DEGs.

3.2 The darkred module, harboring 476 genes, is identified as a key co-expression unit in IPF

Based on scale-free topology and mean connectivity analysis, the optimal soft threshold was determined to be 6 (Figures 3A,B). At this threshold, the network exhibited significant scale-free topology characteristics:the connectivity distribution showed right-skewedness (Supplementary Figure S1), and the double logarithmic coordinate plot showed good fitting (R2 = 0.84, k = −1.2) (Supplementary Figure S2). After merging similar modules, a dendrogram containing 15 distinct gene modules was ultimately obtained (Figure 3C). The gene co-expression modules showed significant correlations with clinical traits (Figure 3D). Furthermore, based on the gene co-expression network, a network heatmap illustrating gene connectivity within each module was generated (Figure 3E). The darkred module (i.e., MEdarkred) showed the most significant correlation with idiopathic pulmonary fibrosis/non-idiopathic pulmonary fibrosis control phenotypes (Figure 3F; Supplementary Table S2). Further analysis revealed a significant positive correlation between gene significance (GS) and module membership (MM) in the darkred module (Figure 3G). Based on these findings, the 476 genes in the darkred module were considered potential IPF-related targets for subsequent analysis.

Construction of the weighted co-expression network related datasets in IPF from the GSE10667 dataset and identification of associated modules. (A,B) Network topology analysis for various soft thresholds. The left panel shows the scale-free fit index (scale-free, y-axis) as a function of the soft threshold power (x-axis); the right panel shows the mean connectivity (degree, y-axis) as a function of the soft threshold power (x-axis). (C) Gene dendrogram obtained by average linkage hierarchical clustering. The colored row below the dendrogram indicates the module assignments determined by the dynamic tree cut method. (D) Heatmap visualization of the clustering results of samples and phenotypic information. (E) Module-trait relationship analysis plot for 15 modules. Each row in the heatmap corresponds to an ME, and each column corresponds to a clinical feature. Each cell contains the corresponding correlation and p-value. (F) Network heatmap of correlations among module genes. (G) Scatter plot of IPF GS and MM in the dark red module.

3.3 Identification of 57 core therapeutic targets for JFG against IPF

Based on the TCMSP and SwissTargetPrediction databases, 1,077 predicted targets of JFG active ingredients were identified. Intersection with IPF DEGs and WGCNA key module genes yielded 57 overlapping genes (Figure 4A; Supplementary Table S3). Protein–protein interaction (PPI) network analysis revealed targets with high degree values as potentially critical nodes (Figure 4B; Tables 1, 2). The study constructed a component-target-disease network, elucidating the multi-target regulatory potential of JFG against IPF (Figures 4C,D). Ultimately, these 57 core targets were screened for subsequent functional validation.

Prediction of potential targets of JFG in IPF treatment and construction of PPI and drug-disease network. (A) Venn diagram of potential target prediction. (B) Original PPI network of intersecting targets. (C) Visualized PPI network of intersecting targets. (D) JFG-IPF network diagram. The red polygon represents IPF. Triangles represent the active components of JFG. Pink circles represent the intersecting targets. Connecting lines indicate the associations between nodes.

Full target nameTarget acronymDegreeBetweennessClosenessFibronectin 1FN142650.58490.55813956Matrix metallopeptidase 9MMP938616.080930.54545456DNA topoisomerase II alphaTOP2A30245.223820.46601942TIMP metallopeptidase inhibitor 1TIMP130107.546820.46153846Matrix metallopeptidase 1MMP12480.0968250.46153846Baculoviral IAP repeat containing 5BIRC530228.928570.45714286Cyclin A2CCNA228166.880950.4528302Matrix metallopeptidase 14MMP142058.4809530.4528302Secreted phosphoprotein 1SPP13278.1531750.44036698Tubulin beta 3 class IIITUBB314144.685710.42857143Alanyl aminopeptidaseANPEP85740.42857143Plasminogen activator, urokinasePLAU2613.932540.42105263Interleukin 10IL102423.90.42105263Matrix metallopeptidase 3MMP3246.8111110.42105263Matrix metallopeptidase 7MMP72274.107140.41379312Matrix metallopeptidase 13MMP13221.44444440.41379312Matrix metallopeptidase 10MMP101812.4761910.41025642Thymidylate synthetaseTYMS3080.947620.40677965CD163 moleculeCD16316102.50.4GLI family zinc finger 1GLI188.5714280.3966942

Information on the top 20 targets sorted by degree value.

Ingredient codeIngredient nameDegreeBetweenness centralityCloseness centralityQianhuQuercetin440.0985173380.417142857JingjieLuteolin160.0305818590.323485968ZhikeBeta-sitosterol120.0398902230.317851959JingjieStigmasterol80.0075391930.289299868GancaoKaempferol80.015902940.380869565ZhikeNobiletin60.0189971020.345971564QianhuSitosterol60.0901898890.358428805GancaoIsorhamnetin60.0159829710.364392679Qianhu(+)-Anomalin40.0086658320.326378539QianhuAmmidin40.0054544940.271375465QianhuIsoimperatorin40.0054544940.271375465GancaoLupiwighteone40.0035211520.375643225GancaoGlyasperin C40.0035211520.375643225GancaoKanzonols w40.0035211520.375643225GancaoGlepidotin A