Integrative Network Toxicology Reveals Potential Molecular Targets Linking Plasticizer Exposure to Inflammatory Gastrointestinal Disorders

Toxicity predictions for the compound structural models of DMP, DEP, DOP, and ATBC were conducted using the ADMETlab 3.0 and ProTox-II databases to obtain relevant information regarding their induced toxicity [ 17 , 18 ] ( Supplementary File S1 ). The SMILES sequences for DMP, DEP, DOP, and ATBC were sourced from the PubChem database. For this study, target prediction was conducted through the ChEMBL [ 19 ], PubChem, and SwissTargetPrediction [ 20 ] databases for target prediction, with the species restricted to Homo sapiens . After merging and removing duplicate entries from the predicted results, all unique targets were retained for subsequent analyses. The UniProt standardization process was then applied, using the IDMapping service to convert target gene names from different databases into UniProtKB identifiers. Subsequently, the UniProt IDMapping service was employed again for batch conversion, yielding a standardized and unified Gene Symbol. Finally, a rigorously validated standardized target library was established for subsequent analyses, ensuring the reliability and reproducibility of the research findings. Common digestive diseases were classified into three categories: inflammatory bowel disease (IBD), esophagitis, and gastritis. IBD included CD and UC. The disease terms were input into the GeneCards [ 21 ], DrugBank [ 22 ], and OMIM [ 23 ] databases to identify relevant targets. For disease-related target collection, only targets with a relevance score greater than 50 were retained from the GeneCards database, whereas all targets obtained from the DrugBank and OMIM databases were included in the subsequent analyses. Subsequently, the UniProt standardization process was applied, using the IDMapping service to convert target gene names obtained from two databases into UniProtKB identifiers (selecting Homo sapiens entries). The predicted targets were merged, and duplicates were removed to construct a standardized target database for CD, UC, esophagitis, and gastritis. Next, a Venn diagram was employed to independently identify common targets between the toxic compounds DMP, DEP, DOP, ATBC and CD, UC, esophagitis, and gastritis. The UniProtKB identifiers of the common targets were uniformly converted to gene names (selecting Homo sapiens entries) for subsequent analyses. These intersecting genes were considered potential targets through which two toxic compounds may induce digestive system toxicity. The intersection genes identified as being involved in the digestive system toxicity induced by DMP, DEP, DOP, and ATBC were input into the STRING [ 24 ] database. The species was restricted to Homo sapiens, and the minimum required interaction score was set to “medium confidence > 0.4” for analysis to obtain the protein–protein interaction (PPI) network diagram. These results were imported into Cytoscape 3.10.4, hub genes were identified using the cytoHubba plugin in Cytoscape with default parameter settings. The central targets of DMP, DEP, DOP, and ATBC were derived from the important nodes obtained by three different topological analysis methods, including degree centrality (DC), betweenness centrality (BC), and closeness centrality (CC) [ 25 ]. Typically, the top fifteen ranked nodes from each method were selected for intersection analysis. To elucidate the core mechanisms and pathways associated with the potential targets related to respiratory toxicity induced by DMP, DEP, DOP, and ATBC, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were conducted [ 26 ]. Functional annotation and pathway analysis were performed using the DAVID database, with Homo sapiens designated as the reference species. GO analysis covered one-dimensional biological process (BP), to clarify the primary biological functions [ 27 ]. KEGG enrichment analysis was used to identify significant pathways related to the potential targets of respiratory toxicity caused by these toxic compounds. Results with statistical significance (adj. p < 0.05) were prioritized for analysis, with the top ten enriched KEGG pathways and GO terms subsequently visualized through bioinformatics tools to facilitate interpretation and visualization of the analytical outcomes. To improve the reliability of core toxic target screening, transcriptomic datasets related to digestive inflammatory diseases were downloaded from the GEO database for validation, including GSE24287 [ 28 ], GSE59071 [ 29 ], and GSE36807 [ 30 ] for CD; GSE13367 [ 31 ], GSE24287 , and GSE179285 [ 32 ] for UC; GSE58640 [ 33 ], GSE190027 [ 34 ], and GSE234973 [ 35 ] for esophagitis; and GSE60427 [ 36 ], GSE47797 [ 37 ], and GSE60662 [ 38 ] for gastritis. To reduce differences among datasets, all differential expression results were standardized using gene symbol, log 2 FC, adjusted p -value(adj. p ), and dataset label, and DEGs were screened with the same criteria: adj. p < 0.05 and |log 2 FC| ≥ 1. For hub target identification, the top 15 targets ranked by degree centrality, betweenness centrality, and closeness centrality were selected from the PPI networks, and their overlapping targets were retained to reduce the bias of a single topological parameter. These hub targets were then cross-analyzed with DEGs from the corresponding GEO datasets, and targets that were both topologically important and significantly differentially expressed were defined as core toxic targets. Bar–volcano plots were generated using ggplot2 to visualize the expression patterns of these core targets. Molecular docking was performed to investigate the interactions between DMP, DEP, DOP, ATBC and their corresponding toxic targets. The main software packages employed included PyMOL (v 3.0.5) [ 39 ] and Chimera(v 1.19) [ 40 ]. High-resolution crystal structures of protein targets were retrieved from the pubchem and Zinc15 database. Chimera was used to remove water molecules and original ligands from protein structures. Subsequently, AutoDock was applied to hydrogenate the receptors (proteins) and perform molecular docking with small-molecule ligands. Binding activities were evaluated on the basis of binding energy. Finally, PyMOL was utilized to analyze and visualize the molecular docking results. MD simulation is an approach used to evaluate the stability and dynamic interactions of proteins and/or between a protein and their ligands. Here, the GROMACS (v2025) MD simulation package was employed to conduct the MD simulation of proteins and their ligands. Protein topologies and the parameters were generated with the AMBER99SB force field, whereas ligand topologies and the parameters were obtained via the Chimera server. In this study, the structural properties evaluated were RMSD [ 41 ], radius of gyration (Rg) [ 42 ], solvent-accessible surface area (SASA) [ 43 ], RMSF [ 44 ], hydrogen bonding, secondary structure, and angular and distance measurements. Simulation trajectories were performed using CMD (v10.0.26100.2894), and plots were generated with DuIvyTools (v0.5.0) ( https://duivytools.readthedocs.io/ ; accessed on 27 February 2026). Here, we selected the single-cell dataset GSE214695 (comprising 6 control samples, 6 CD colon samples, and 6 UC colon samples) [ 45 ], GSE254513 (comprising 12 control samples and 15 samples of gastritis) [ 46 ], GSE201153 (comprising 5 control samples and 5 samples of esophagitis) [ 47 ] for analysis. We performed rigorous data curation and downstream analysis of 46,700 single cells using Seurat R package (v5.3.0). Samples were then pooled together in the same object. Low-quality cells were then filtered out based on mitochondrial RNA percentage and number of genes per cell. Immunoglobulin (IG) genes were removed from all of the main cell types except B and plasma cells to reduce background noise. Dimensionality reduction was conducted by using the Uniform Manifold Approximation and Projection (UMAP) algorithm using the optimal number of PCs. Marker genes were utilized to define each subcluster within the main cell types. The FindAllMarkers function was used to identify marker genes using the default threshold parameters, excluding the min.pct and the thresh.use, both of which were set to 0.25. All statistical analyses were performed using R software (version 4.4.2). For transcriptomic datasets obtained from GEO, differential expression analysis was performed between disease and control groups using the limma package. The moderated t -test implemented in limma was used to estimate differential expression, and p -values were adjusted for multiple testing using the Benjamini–Hochberg false discovery rate (FDR) method. Differentially expressed genes were defined as those with adj. p < 0.05 and |log 2 FC| ≥ 1. For functional enrichment analyses, GO and KEGG terms with adj. p < 0.05 were considered statistically significant. For single-cell RNA-seq analysis, marker genes or differentially expressed genes among cell clusters were identified using the Wilcoxon rank-sum test, where applicable. Unless otherwise stated, adj. p < 0.05 was considered statistically significant.

Plasticizers are chemical additives widely used to improve the flexibility, durability, and processability of plastic materials. Among them, phthalate esters, including dimethyl phthalate (DMP), diethyl phthalate (DEP), and dioctyl phthalate/di(2-ethylhexyl) phthalate (DOP/DEHP), represent some of the most extensively used plasticizers. Acetyl tributyl citrate (ATBC), a non-phthalate plasticizer, has also been increasingly applied as an alternative in food-contact materials, medical products, personal care products, and other consumer goods [ 1 , 2 ]. ATBC is a phthalate-free plasticizer widely used as an alternative to conventional phthalates in food-contact materials, medical devices, toys, and consumer products. ATBC does not belong to the phthalate ester family. Phthalates possess an aromatic benzene ring as the core structure, while ATBC is an aliphatic citrate ester without aromatic rings. Although both contain ester bonds, their parent skeletons, functional groups and chemical classifications are distinctly different. All four compounds are widely used as plasticizers for plastics, so they are often compared in the same application scenarios. Because plasticizers are not covalently bound to polymer matrices, they can migrate from plastic products into food, water, dust, and biological systems, resulting in continuous human exposure through ingestion, inhalation, dermal absorption, and medical contact [ 3 ]. Growing evidence suggests that plasticizers and their related compounds may pose substantial risks to human health [ 4 , 5 ]. Phthalates and their metabolites have been associated with endocrine disruption, reproductive toxicity, metabolic abnormalities, immune dysregulation, oxidative stress, and inflammatory injury. The gastrointestinal tract is particularly vulnerable to plasticizer exposure because it represents a major route of chemical intake and a direct interface between environmental toxicants, intestinal epithelial cells, gut microbiota, and mucosal immune responses. Long-term exposure to plasticizers may impair epithelial barrier integrity, alter tight junction function, disturb gut microbiota homeostasis, and activate inflammatory signaling pathways, thereby contributing to chronic mucosal injury and gastrointestinal dysfunction [ 6 ]. Inflammatory gastrointestinal disorders (IGDs) are complex diseases driven by interactions among genetic susceptibility, immune dysregulation, environmental exposure, microbial imbalance, and epithelial barrier dysfunction [ 7 ]. Epidemiological studies indicate that inflammatory gastrointestinal diseases represent a substantial and increasing global health burden. For example, the global prevalence of ulcerative colitis was estimated to be approximately five million cases in 2023, and the incidence of inflammatory bowel disease has continued to rise in many regions worldwide. GERD-related esophageal inflammatory conditions and gastritis/duodenitis are also highly prevalent upper gastrointestinal disorders and contribute substantially to global digestive disease burden. Crohn’s disease (CD) and ulcerative colitis (UC), the two major forms of inflammatory bowel disease, are characterized by chronic relapsing intestinal inflammation, abdominal pain, diarrhea, and progressive tissue damage [ 8 ]. Esophagitis and gastritis are common upper gastrointestinal inflammatory disorders involving epithelial injury, immune-cell infiltration, cytokine activation, and tissue remodeling [ 9 , 10 , 11 ]. Given that plasticizer exposure can promote oxidative stress, barrier disruption, immune activation, and microbiota disturbance, these compounds may be associated with the initiation or progression of CD, UC, esophagitis, and gastritis. Several studies have begun to link plasticizer exposure with disease progression. Phthalates have been implicated in inflammatory, metabolic, endocrine, reproductive, and tumor-related disorders, suggesting broad systemic toxicity. In gastrointestinal contexts, experimental studies indicate that phthalate exposure may aggravate intestinal inflammation by disrupting epithelial barrier function, reshaping gut microbiota composition, and enhancing pro-inflammatory mediator production [ 12 ]. Network toxicology studies have also suggested that plasticizers may affect disease progression by interacting with key proteins and signaling pathways. However, current evidence remains fragmented, and most studies have focused on individual compounds, isolated toxic endpoints, or single disease models [ 12 , 13 , 14 , 15 , 16 ]. The potential effects of DMP, DEP, DOP/DEHP, and ATBC on CD, UC, esophagitis, and gastritis have not been systematically compared. Therefore, the molecular mechanisms linking specific plasticizers to distinct IGDs remain insufficiently defined. In particular, it remains unclear whether different plasticizers converge on shared inflammatory pathways or act through disease-specific molecular targets. It is also unknown which candidate targets may mediate plasticizer-associated toxicity across lower and upper gastrointestinal inflammatory diseases. To address these gaps, the present study established an integrative network toxicology framework to systematically investigate the potential links between four representative plasticizers—DMP, DEP, DOP/DEHP, and ATBC—and four IGDs, including CD, UC, esophagitis, and gastritis. First, putative plasticizer-associated targets and disease-related targets were collected from public databases and integrated to identify overlapping toxicant–disease targets. Protein–protein interaction network analysis, topological screening, and functional enrichment analysis were then performed to prioritize key targets and perturbed biological pathways. These candidates were further refined using transcriptomic differential expression analysis across disease-relevant GEO datasets. Finally, molecular docking and molecular dynamics simulations were applied to evaluate the binding affinity and structural stability between representative plasticizers and prioritized targets. Single-cell RNA-seq analysis was then used to characterize the cell-type-specific expression patterns of key targets in diseased tissues. Through this multi-step strategy, this study aimed to identify disease-specific and plasticizer-associated molecular targets and provide mechanistic insights into how plasticizer-related compounds may contribute to inflammatory gastrointestinal disease progression.

In this study, the PubChem database was used to search for DMP, DEP, DOP, and ATBC to obtain their standard structures and SMILES sequences ( Table S1 in Supplementary File S2 ). The systemic and organ-specific toxicity profiles of DMP, DEP, DOP, and ATBC were predicted using ADMETlab 3.0 and ProTox-II databases. Both platforms consistently indicated that these four compounds exhibit toxicity towards gastrointestinal organs, especially highlighting the potential for liver injury. Additional toxic endpoints predicted by consensus included ocular irritation, skin sensitization, hepatotoxicity, nephrotoxicity, and neurotoxicity. The ProTox-II database classified the overall toxicity levels of DMP, DEP, DOP, and ATBC as Grade 6, Grade 6, Grade 4, and Grade 4, respectively. Their predicted human oral median lethal doses (LD50s) were 6800 mg/kg, 6172 mg/kg, 1340 mg/kg, and 517 mg/kg, respectively. Potential targets for DMP (398), DEP (593), DOP (272), and ATBC (484) were identified by integrating data from the ChEMBL, PubChem, and SwissTargetPrediction databases, followed by the removal of duplicate entries. Concurrently, CD-associated targets (n = 2908) were obtained from the GeneCards, DrugBank, and OMIM databases. Venn diagram analysis identified 166, 249, 119, and 234 overlapping CD-related targets for DMP, DEP, DOP, and ATBC, respectively ( Figure 1 A–D). A disease–toxic compound–target network was then constructed ( Figure 1 E), and PPI networks of the intersecting targets were generated ( Supplementary File S2, Figure S1 ). GO and KEGG enrichment analyses were performed separately for the overlapping targets of each plasticizer–CD pair ( Figure 1 F–I). DMP–CD targets were mainly enriched in response to xenobiotic stimulus and measles pathways; DEP–CD targets were associated with response to xenobiotic stimulus and lipid and atherosclerosis pathways; DOP–CD targets were related to the ERK1/2 cascade and chemokine signaling pathway; and ATBC–CD targets were enriched in positive regulation of the MAPK cascade and MAPK signaling pathway. Subsequently, Cytoscape 3.8.2 was used to analyze the topological properties of the PPI networks, including degree centrality (DC), betweenness centrality (BC), and closeness centrality (CC). The top 15 genes ranked by each parameter were selected, and their intersections were visualized using Venn diagrams ( Supplementary File S2, Figure S2 ). The intersecting hub genes were then used to construct network diagrams ( Figure 1 J–M). Finally, the four sets of hub targets were merged, deduplicated, and intersected with DEGs from the GSE24287 , GSE59071 , and GSE36807 datasets. This analysis identified two significantly differentially expressed core targets in CD, CXCL8 , and FN1 ( Figure 1 N). Venn diagram analysis identified 155, 234, 114, and 227 potential overlapping targets between DMP, DEP, DOP, and ATBC in UC, respectively ( Figure 2 A–D). The corresponding TTD network is presented in Figure 1 E. Meanwhile, a PPI network of the intersecting targets was constructed ( Figure S3 in Supplementary File S2 ). Subsequently, the overlapping targets of DMP, DEP, DOP, and ATBC with UC were separately subjected to GO analysis and KEGG pathway analysis ( Figure 2 F–I). These analyses revealed that DMP-UC targets were mainly enriched in response to xenobiotic stimulus and Measles. DEP-UC targets were associated with response to xenobiotic stimulus and lipid and atherosclerosis. DOP-UC targets were mainly enriched in response to xenobiotic stimulus and Chemokine signaling pathway. ATBC-UC targets were mainly involved in positive regulation of kinase activity and MAPK signaling pathway. Cytoscape 3.8.2 was used to analyze the topological properties of the PPI networks, including DC, BC, and CC, the top 15 core genes were selected to generate a Venn diagram ( Figure S4 in Supplementary File S2 ), PPI network analysis was conducted as described for CD ( Figure 2 J–M). Following merging and intersecting with DEGs from the GSE13367 , GSE24287 , and GSE179285 datasets, one core target showing significant differential expression in UC was identified: IL1B ( Figure 2 N). Venn diagram analysis identified 274, 386, 190, and 383 potential overlapping targets for DMP, DEP, DOP, and ATBC in esophagitis, respectively ( Figure 3 A–D). A TTD network was shown in Figure 3 E. Meanwhile, a PPI network of the intersecting targets was constructed ( Figure S5 in Supplementary File S2 ). The overlapping targets of the four compounds with esophagitis were separately subjected to GO analysis and KEGG pathway analysis ( Figure 3 B–F). These analyses showed that the targets of DMP and DEP in esophagitis were mainly enriched in response to xenobiotic stimulus and neuroactive ligand–receptor interaction. Targets of DOP in esophagitis were mainly associated with response to xenobiotic stimulus and morphine addiction. Targets of ATBC in esophagitis were mainly enriched in positive regulation of transferase activity and MAPK signaling pathway. Subsequently, Cytoscape 3.8.2 was applied to analyze the topological properties of the PPI networks, including DC, BC, and CC, the top 15 core genes were selected to generate a Venn diagram ( Figure S6 in Supplementary File S2 ) and the topological properties of the PPI networks were analyzed and visualized ( Figure 3 J–M). The four sets of overlapping targets were then merged and deduplicated, followed by intersection analysis with DEGs from the GSE58640 , GSE190027 , and GSE234973 datasets. Seven core targets showing significant differential expression in esophagitis were identified: MAPK3 , FASN , FN1 , PPARG , CXCL8 , FOS , and HIF1A . The screening results were visualized using multiple grouped bar–volcano plots ( Figure 3 N). A total of 274, 386, 190, and 383 potential overlapping targets were identified for DMP, DEP, DOP, and ATBC in gastritis, respectively ( Figure 4 A–D). A TTD network was constructed and presented in Figure 4 E. Meanwhile, a PPI network of the intersecting targets was constructed ( Figure S7 in Supplementary File S2 ). The overlapping targets of DMP, DEP, DOP, and ATBC with gastritis were separately subjected to GO analysis and KEGG pathway analysis ( Figure 4 F–I). Enrichment analysis revealed distinct pathways: Targets of DMP in gastritis were mainly enriched in response to molecules of bacterial origin and lipid/atherosclerosis pathways. Targets of DEP were mainly associated with response to lipopolysaccharide and lipid/atherosclerosis. DOP targets were mainly involved in cell homeostasis and lipid/atherosclerosis. Targets of ATBC were mainly enriched in positive regulation of kinase activity and the chemokine signaling pathway. Then, Cytoscape 3.8.2 was used to analyze the topological properties of the PPI networks, including DC, BC, and CC, the top 15 core genes were selected to generate a Venn diagram ( Figure S8 in Supplementary File S2 ), and the intersecting genes were used to construct a network diagram ( Figure 4 J–M). Integration with DEGs from the GSE60427 , GSE47797 , and GSE60662 datasets identified seven core targets with significant differential expression in gastritis: MMP9 , TNF , TLR4 , IL6 , CCR2 , IFNG , and PTGS2 ( Figure 4 N). We determined the common differentially expressed target genes of DMP in CD, UC, esophagitis, and gastritis by taking the intersection. MMP7 was identified as the sole target consistently dysregulated in three or more of these conditions ( Figure 5 A). Functional enrichment analyses (GO, KEGG, and Disease Ontology) indicated that the shared targets were significantly enriched in biological processes related to response to xenobiotic stimulus and pathways associated with bladder cancer ( Figure 5 B). Moreover, these targets exhibited strong associations with various inflammatory and organ-specific diseases, including rheumatoid arthritis, cholesteatoma, cholesteatoma of middle ear, viral hepatitis, hepatobiliary diseases, acute kidney failure, chronic liver diseases, and non-small cell lung cancer ( Figure 5 C). To further explore the binding affinity between MMP7 and DMP, molecular docking simulations ( Figure 5 D) and 100-ns-atom molecular dynamics simulations ( Figure 5 E–H) were performed, confirming stable binding and favorable binding affinity. Shared differential targets of DEP in at least three IGDs were identified and are visualized using a Venn diagram ( Figure 6 A). HMOX1 and NOS2 were found to be the only two targets consistently dysregulated in three or more of the four IGDs. Subsequently, GO functional enrichment, KEGG pathway enrichment ( Figure 6 B), and Disease Ontology (DO) analysis ( Figure 6 C) were conducted on these common differential targets. These targets were mainly enriched in response to lipopolysaccharide and Chagas disease, and were associated with a spectrum of diseases including asthma, bronchial diseases, endocrine system diseases, intestinal diseases, liver cirrhosis, lung diseases, pancreas diseases, pancreatitis, urinary bladder cancer, and urinary system cancers. To characterize the binding affinity between DEP and its top candidate targets, molecular docking ( Figure 6 D,E) and molecular dynamics simulations ( Figure 6 F–I) were further performed. Collectively, these results indicate that DEP forms a thermodynamically stable complex with HMOX1, exhibiting favorable binding affinity and sustained intermolecular interactions throughout the simulation trajectory. Notably, DOP does not share any common target molecules in at least three IGDs, and this has been visually demonstrated using a Venn diagram ( Figure 7 A). Consequently, GO functional enrichment, KEGG pathway enrichment ( Figure 7 B), and DO analyses ( Figure 7 C) were performed exclusively on this set of shared targets. The results indicated significant association with biological processes such as cell chemotaxis and phospholipase D signaling pathway. Moreover, DO analysis revealed statistically enriched associations with a spectrum of non-GD conditions, including asthma, breast carcinoma in situ, ductal carcinoma in situ, lipid storage disease, lung disease, nasal cavity disease, pre-malignant neoplasm, rhinitis, sarcoidosis, and upper respiratory tract disease. Shared differential targets of DEP in at least three IGDs were identified and are visualized using a Venn diagram ( Figure 8 A). LTF and CCL11 were identified as the two targets commonly dysregulated across three or more diseases. Functional characterization of these common targets was subsequently performed via GO functional enrichment, KEGG pathway enrichment ( Figure 8 B), and DO enrichment analysis ( Figure 8 C). GO analysis revealed significant enrichment in biological processes, including positive regulation of transferase activity and apoptosis. DO analysis revealed that the common targets associated with a spectrum of pathologies, including asthma, bronchial disease, carcinoma, cell type cancer, endometriosis, hypersensitivity reaction disease, interstitial lung disease, lung disease, periodontitis, and urinary system cancer. To investigate the binding affinity between ATBC and the top candidate targets, molecular docking ( Figure 8 D,E) and molecular dynamics simulations ( Figure 8 F–I) were further performed. These computational analyses demonstrate that CCL11 exhibits robust and stable binding affinity toward ATBC, suggesting its potential role as a key mechanistic mediator in ATBC’s therapeutic effects. Single-cell RNA sequencing (scRNA-seq) analyses were performed across three independent inflammatory gastrointestinal disease cohorts to characterize the cellular context-specific expression of key disease-associated targets and key toxicant-associated targets identified in prior network pharmacology and toxicity screening steps. First, scRNA-seq data from the GSE214695 dataset, comprising six healthy control samples, six CD patient samples, and six UC patient samples, were processed and clustered. Cell-type annotation revealed conserved major lineages, such as epithelial cells, T cells, macrophages, and stromal cells, across samples. Corresponding cell type distribution maps for CD and control groups are shown in Figure 9 A. Heatmap visualization of target gene expression across annotated cell types demonstrated that FN1 exhibited extremely high expression in stromal cells in CD tissues ( Figure 9 B), a finding consistent with its established role in extracellular matrix remodeling and fibro-inflammatory activation. Parallel analysis of UC versus control samples ( Figure 9 C) revealed broadly attenuated expression of all signature targets across immune and structural cell populations, as summarized in Figure 9 D. Second, scRNA-seq data from the gastritis cohort ( GSE254513 , 12 controls, 15 gastritis samples) were analyzed. Cell-type composition maps for gastritis and control samples are presented in Figure 9 E. Heatmap-based expression profiling identified IFNG as significantly enriched in NK cells ( Figure 9 F), suggesting a potential NK-mediated immunomodulatory axis in gastric mucosal inflammation. Third, scRNA-seq data from the esophagitis cohort ( GSE201153 , five controls, five esophagitis samples) were integrated and annotated. Cell type distribution maps for esophagitis and control samples are displayed in Figure 9 G. As shown in Figure 9 H, FOS , a core immediate early transcription factor, exhibited broad, high-level expression across various epithelial and immune cell types ( Figure 9 H), indicating widespread transcriptional activation in response to esophageal injury or inflammation. Collectively, these cross-disease, cell-type specific analyses delineate distinct functional contexts for the prioritized targets, supporting their biological relevance and informing hypothesis-driven mechanistic validation.

Plasticizers constitute a class of chemical additives extensively employed in polymeric materials to enhance processability, flexibility, and mechanical durability. By reducing intermolecular forces within polymer matrices, they improve material pliability and impact resilience, thereby expanding the utility of plastic products across diverse sectors, including medical devices, food packaging, and personal care products [ 48 , 49 ]. The widespread application of these additives has led to pervasive environmental and subsequent human exposure via inhalation and ingestion [ 49 ]. In this study, we established an integrative network toxicology framework to systematically investigate the potential molecular links between four representative plasticizers—DMP, DEP, DOP/DEHP [ 1 , 50 ], and ATBC [ 51 , 52 ]—and four IGDs, including CD, UC, esophagitis, and gastritis. By integrating target prediction, disease-associated gene collection, PPI network analysis, transcriptomic validation, molecular docking, molecular dynamics simulation, and single-cell RNA-seq analysis, we identified both disease-specific and plasticizer-associated candidate targets. These findings suggest that plasticizer-related compounds may influence gastrointestinal inflammation through multiple biological processes, including xenobiotic response, chemokine signaling, MAPK-related stress responses, extracellular matrix remodeling, epithelial injury, and immune-cell activation. [ 53 , 54 , 55 , 56 , 57 ]. Our disease-specific analysis identified CXCL8 and FN1 as candidate targets associated with CD, IL1B with UC, MAPK3 , FASN , FN1 , PPARG , CXCL8 , FOS , and HIF1A with esophagitis, and MMP9 , TNF , TLR4 , IL6 , CCR2 , IFNG , and PTGS2 with gastritis. These targets are biologically consistent with the known pathological features of inflammatory gastrointestinal diseases. CXCL8 is a central pro-inflammatory chemokine involved in neutrophil recruitment and mucosal immune activation and has been reported in CD, UC, gastritis, and esophagitis [ 58 , 59 , 60 , 61 ]. This is consistent with our finding that CXCL8 may represent a shared signature target across several IGDs. Notably, a previous clinical study involving pediatric patients with eosinophilic esophagitis (EoE) demonstrated that CXCL8 is one of the few significantly upregulated biomarkers in active EoE compared with healthy controls, and its elevation is strictly dependent on ex vivo immune stimulation [ 58 ]. FN1 , an extracellular matrix glycoprotein, may reflect stromal activation, tissue remodeling, and fibro-inflammatory responses, particularly in CD [ 62 , 63 , 64 ]. IL1B is a key inflammasome-related cytokine and an important mediator of intestinal inflammation in IBD [ 65 ]. In esophagitis, MAPK3 and FOS may indicate stress-response and inflammatory transcriptional activation, whereas HIF1A and PPARG may participate in epithelial injury, metabolic adaptation, and inflammatory regulation [ 66 , 67 ]. In gastritis, TNF , TLR4 , IL6 , IFNG , PTGS2 , MMP9 , and CCR2 represent inflammatory cytokine signaling, innate immune activation, monocyte recruitment, and mucosal tissue remodeling [ 68 , 69 , 70 , 71 , 72 ]. Together, these findings indicate that plasticizer-associated molecular perturbations may converge on inflammatory and remodeling pathways that are central to gastrointestinal disease progression. These findings are partly consistent with previous experimental studies on plasticizer-induced gastrointestinal toxicity. Increasing evidence suggests that chronic or repeated exposure to plasticizers may compromise intestinal epithelial barrier function, alter gut microbiota composition, induce oxidative stress, and trigger inflammatory signaling pathways, thereby contributing to intestinal toxicity and gastrointestinal dysfunction [ 73 , 74 , 75 ]. In addition, although ATBC is widely used as a phthalate-free alternative plasticizer, recent evidence suggests that chronic ATBC exposure may also induce intestinal toxicity and inflammatory responses. This supports our finding that ATBC-associated targets, such as LTF and CCL11 , may be involved in mucosal immune regulation and inflammatory processes. However, the current literature remains limited, particularly regarding long-term human exposure, dose–response relationships, and direct evidence linking ATBC exposure to inflammatory gastrointestinal diseases. Therefore, our computational results are consistent with some reported toxicological effects, but they should be interpreted as mechanistic hypotheses rather than definitive causal evidence. Cross-disease toxicant-specific analysis further identified MMP7 as a DMP-associated target, HMOX1 and NOS2 as DEP-associated targets, and LTF and CCL11 as ATBC-associated targets. These targets suggest that different plasticizers may affect IGDs through partially distinct mechanisms. MMP7 is involved in extracellular matrix degradation, epithelial repair, and inflammatory tissue remodeling, suggesting that DMP may be linked to mucosal remodeling processes [ 76 ]. HMOX1 and NOS2 are closely related to oxidative stress, nitric oxide metabolism, and inflammatory responses, supporting a potential role of DEP in redox imbalance and immune activation [ 77 ]. LTF and CCL11 are associated with antimicrobial defense, epithelial immunity, eosinophil recruitment, and allergic-type inflammation, suggesting that ATBC may be connected with mucosal immune modulation [ 78 ]. Notably, DOP/DEHP did not show a common target shared across three or more diseases, implying that its potential effects may be more disease-specific or mediated through heterogeneous molecular pathways rather than a single convergent target. Molecular docking and molecular dynamics simulations provided additional structural-level support for the predicted toxicant–target relationships. The stable binding patterns observed between selected plasticizers and their candidate targets suggest that these compounds may have the potential to interact with key inflammatory or remodeling-related proteins. However, these results should be interpreted as computational evidence of binding feasibility rather than direct proof of toxicological causality. Docking and dynamics simulations cannot determine whether plasticizers regulate gene expression, protein activity, or downstream signaling in biological systems. Therefore, the predicted interactions require further validation using cellular models, animal experiments, and exposure-response assays. The single-cell RNA-seq analysis further strengthened the biological relevance of the prioritized targets by revealing their cell-type-specific expression patterns in diseased tissues. In CD tissues, FN1 showed high expression in stromal cells, supporting its role in extracellular matrix remodeling and fibro-inflammatory activation. In gastritis, IFNG was enriched in NK cells, suggesting that NK-cell-mediated immune responses may contribute to gastric mucosal inflammation. In esophagitis, FOS was broadly expressed across epithelial and immune cell populations, indicating widespread transcriptional activation under inflammatory or injury-related conditions. These observations provide cellular context for the network toxicology findings and suggest that plasticizer-associated targets may act within specific mucosal microenvironmental compartments, including stromal, epithelial, and immune-cell niches. This study has several strengths. First, we established an integrative network toxicology framework combining public target prediction, disease-related target collection, PPI network analysis, topological screening, functional enrichment analysis, transcriptomic validation, molecular docking, molecular dynamics simulation, and single-cell RNA-seq analysis. This multi-step strategy allowed us to systematically identify disease-specific and plasticizer-associated candidate targets across multiple inflammatory gastrointestinal disorders. Second, by integrating multiple GEO datasets and single-cell transcriptomic data, this study provided both transcriptomic and cell-type-specific evidence to support the biological relevance of the predicted targets. Third, we systematically evaluated the potential toxicological mechanisms as well as the similarities and differences of plasticizer exposure in inflammatory gastrointestinal disorders. This study also has several limitations. First, the findings are mainly based on public databases, computational target prediction, and transcriptomic integration; toxicological targets predicted via ChEMBL, PubChem and SwissTargetPrediction and disease-related targets identified using GeneCards, DrugBank and OMIM are all in silico predictions, which may lead to false positive outcomes. By contrast, datasets from the GEO database are derived from experimental evidence. Therefore, they cannot establish direct causal relationships between plasticizer exposure and disease progression. Second, molecular docking and molecular dynamics simulations only evaluate predicted binding affinity and structural stability, but do not confirm actual protein activity changes or downstream pathway activation. Third, the GEO and single-cell datasets used in this study do not contain direct measurements of plasticizer exposure, exposure duration, tissue concentration, or metabolite levels. Fourth, the combined effects of mixed plasticizer exposure were not evaluated, although real-world exposure usually involves multiple compounds simultaneously. Finally, due to limitations in time, funding, and experimental resources, no in vitro or in vivo validation experiments (e.g., ELISA, Western blotting, cell-based assays, or animal studies) were performed in the current study. Therefore, the proposed mechanisms and the potential roles of CXCL8 , FN1 , HMOX1 , NOS2 , LTF , and CCL11 should be considered preliminary and require further experimental validation in future studies. Future studies should further validate these candidate targets and pathways using intestinal and gastric epithelial cell models, immune-cell co-culture systems, and animal models of gastrointestinal inflammation under chronic plasticizer exposure. In addition, population-based studies with direct measurements of plasticizer metabolites, exposure duration, and dose–response relationships are needed to clarify the relevance of these findings in humans. The combined effects of mixed plasticizer exposure should also be investigated, as real-world exposure usually involves multiple compounds simultaneously.

In conclusion, this study provides a systematic computational framework for exploring the potential molecular associations between plasticizer exposure and IGDs. By identifying disease-specific and plasticizer-associated molecular targets and integrating transcriptomic, structural, and single-cell evidence, our findings offer mechanistic clues for understanding how plasticizer-related compounds may be involved in gastrointestinal inflammation. These results may help prioritize candidate biomarkers and molecular targets for future toxicological validation, environmental risk assessment, and prevention strategies for IGDs.