Unveiling the mechanistic links between plasticizers and gastric cancer via network toxicology and molecular docking approaches.

Expression data from GSE65801 and GSE79973 were merged to identify genes related to gastric cancer. Box plots were created from these data sets (Fig.  2 A-B) showing that the batch effect was significantly reduced through PCA, and homogeneity among the samples was achieved (Fig.  2 C- D). In addition, the PVCA analysis results showed that the proportion of variance explained by batch decreased from 81% to 6% (Supplementary Fig.  1 ). Fig. 2 Screening potential genes related to plasticizers and gastric cancer. ( A ) Boxplot comparison of gene expression levels between GSE65801 and GSE79973 datasets before batch correction. The data is divided by projects and samples, with each box representing the distribution of gene expression values for each sample within a given project. ( B ) Boxplot comparison of gene expression levels between GSE65801 and GSE79973 datasets after batch correction. ( C ) The PCA scatter plot shows distinct separation between the GSE65801 and GSE79973 datasets before batch correction, indicating batch effects. ( D ) PCA scatter plot after batch correction shows integration of the GSE65801 and GSE79973 datasets, indicating reduced batch effects. ( E ) Volcano plot shows DEGs based on logFC and significance. Red dots are up-regulated, green dots are down-regulated, and grey dots are non-significant. ( F ) A Venn diagram shows 17 overlapping genes linked to plasticizer exposure (red) and gastric cancer (blue) Screening potential genes related to plasticizers and gastric cancer. ( A ) Boxplot comparison of gene expression levels between GSE65801 and GSE79973 datasets before batch correction. The data is divided by projects and samples, with each box representing the distribution of gene expression values for each sample within a given project. ( B ) Boxplot comparison of gene expression levels between GSE65801 and GSE79973 datasets after batch correction. ( C ) The PCA scatter plot shows distinct separation between the GSE65801 and GSE79973 datasets before batch correction, indicating batch effects. ( D ) PCA scatter plot after batch correction shows integration of the GSE65801 and GSE79973 datasets, indicating reduced batch effects. ( E ) Volcano plot shows DEGs based on logFC and significance. Red dots are up-regulated, green dots are down-regulated, and grey dots are non-significant. ( F ) A Venn diagram shows 17 overlapping genes linked to plasticizer exposure (red) and gastric cancer (blue) Differential analysis identified 1012 genes with altered expression between normal and gastric cancer tissues, comprising 440 up-regulated and 572 down-regulated genes (Fig.  2 E). Using molecular formulas and SMILES strings of BBP, DBP, and DEHP as queries (Supplementary Table 3 ), we consolidated several databases, removed redundancies, and obtained 271 candidate protein targets. The overlap between DEGs and the targeted plasticizers highlighted 17 genes of interest (Fig.  2 F). For the 17 target genes, a comprehensive evaluation of 113 machine learning models identified RF + Lasso as the optimal algorithm for screening plasticizer-associated gastric cancer markers. This model demonstrated robust performance on the independent test set, with an AUC of 0.816, accuracy of 75.3%, and balanced sensitivity (73.7%) and specificity (76.8%) (Fig.  3 A). Notably, the selected model exhibited minimal overfitting, as evidenced by the close alignment between training (AUC: 0.981) and test set performance (AUC: 0.816) (see Supplementary Table 4 for the detailed evaluation metrics of the top ten performers). This rigorous process identified six core genes with strong diagnostic potential: carboxypeptidase B1 (CPB1), aldo-keto reductase family one member C1 (AKR1C1), glutamate metabotropic receptor 2 (GRM2), carbonic anhydrase 2 (CA2), matrix metallopeptidase 7 (MMP7), and tryptophan 2,3-dioxygenase (TDO2). The diagnostic potential of these core genes was verified through ROC curve analysis (Fig.  3 B). Fig. 3 Identification of hub genes in plasticizer-related gastric cancer via machine learning. ( A ) Model Performance Comparison: The bar chart shows the evaluation metrics for different models in the test group. Each color represents a different evaluation metric. The numerical values of the evaluation metrics are set as the X-axis variable. The RF + Lasso (highlighted) was selected based on its superior, balanced performance on the test set (AUC = 0.816). ( B ) ROC Curves: ROC curves for six hub genes (CPB1, AKR1C1, GRM2, CA2, MMP7, and TDO2). X-axis = false positive rate, Y-axis = sensitivity. AUC indicates predictive performance. ( C ) A volcano plot displays the differential expression of the six hub genes, with logFC and significance shown. Red dots are up-regulated, green dots are down-regulated. ( D ) The box plot displays the differential expression of six hub genes in normal and gastric cancer samples. * P  < 0.05, ** P  < 0.01; *** P  < 0.001. ( E ) Correlation analysis of the expression of six hub genes. This figure shows the Pearson correlation coefficients for the expression levels of the six hub genes. Each cell represents the strength and direction of the correlation between two genes. The histogram along the diagonal represents the distribution of individual gene expression values, and the scatterplots show pairwise relationships between genes. Correlation coefficients and significance levels are displayed for each gene pair. * P  < 0.05 Identification of hub genes in plasticizer-related gastric cancer via machine learning. ( A ) Model Performance Comparison: The bar chart shows the evaluation metrics for different models in the test group. Each color represents a different evaluation metric. The numerical values of the evaluation metrics are set as the X-axis variable. The RF + Lasso (highlighted) was selected based on its superior, balanced performance on the test set (AUC = 0.816). ( B ) ROC Curves: ROC curves for six hub genes (CPB1, AKR1C1, GRM2, CA2, MMP7, and TDO2). X-axis = false positive rate, Y-axis = sensitivity. AUC indicates predictive performance. ( C ) A volcano plot displays the differential expression of the six hub genes, with logFC and significance shown. Red dots are up-regulated, green dots are down-regulated. ( D ) The box plot displays the differential expression of six hub genes in normal and gastric cancer samples. * P  < 0.05, ** P  < 0.01; *** P  < 0.001. ( E ) Correlation analysis of the expression of six hub genes. This figure shows the Pearson correlation coefficients for the expression levels of the six hub genes. Each cell represents the strength and direction of the correlation between two genes. The histogram along the diagonal represents the distribution of individual gene expression values, and the scatterplots show pairwise relationships between genes. Correlation coefficients and significance levels are displayed for each gene pair. * P  < 0.05 Volcano and violin plots were employed to depict the expression differences of the six pivotal genes in gastric cancer samples (Fig.  3 C-D). Further correlation analysis demonstrated that CPB1 expression correlated positively with CA2 (OR: 0.34, P  < 0.05), AKR1C1 correlated inversely with MMP7 (OR: -0.32, P  < 0.05), while GRM2 and MMP7 both showed positive correlations with TDO2 (OR: 0.29, P  < 0.05; OR: 0.31, P  < 0.05) (Fig.  3 E). In gastric cancer, GO analysis demonstrated that the six target genes are primarily associated with cellular metabolic processes, with the main cellular components enriched at the synaptic membrane and GABA receptor complex. The primary molecular functions were related to the activity of cellular ion channels (Fig.  4 A). Through KEGG pathway analysis, we found these genes to be primarily enriched in the neural activity ligand signaling and linoleic acid metabolism pathways (Fig.  4 B). Fig. 4 Functional enrichment analysis of six hub genes and external data validation. ( A ) GO enrichment annotates overlapping genes in BP, CC, and MF. X-axis = gene ratio, color gradient = p-value (red = higher significance). ( B ) KEGG analysis shows enriched pathways for overlapping genes. X-axis = gene ratio, dot size = gene count, color gradient = p-value (red = higher significance). ( C ) A box plot shows the differential expression of six hub genes (CPB1, AKR1C1, GRM2, CA2, MMP7, and TDO2) in STAD and normal samples. * P  < 0.05, ** P  < 0.01; *** P  < 0.001. ( D ) ROC Curves: ROC curves for six hub genes in STAD. X-axis = false positive rate, Y-axis = sensitivity. AUC indicates predictive performance Functional enrichment analysis of six hub genes and external data validation. ( A ) GO enrichment annotates overlapping genes in BP, CC, and MF. X-axis = gene ratio, color gradient = p-value (red = higher significance). ( B ) KEGG analysis shows enriched pathways for overlapping genes. X-axis = gene ratio, dot size = gene count, color gradient = p-value (red = higher significance). ( C ) A box plot shows the differential expression of six hub genes (CPB1, AKR1C1, GRM2, CA2, MMP7, and TDO2) in STAD and normal samples. * P  < 0.05, ** P  < 0.01; *** P  < 0.001. ( D ) ROC Curves: ROC curves for six hub genes in STAD. X-axis = false positive rate, Y-axis = sensitivity. AUC indicates predictive performance To enhance the reliability of the observed expression differences among the six hub genes, mRNA sequencing data from the TCGA database were further analyzed to validate their differential expression between normal and tumor samples. Consistent with the findings mentioned above, GRM2, MMP7, and TDO2 exhibited significantly higher expression levels in gastric cancer tissues, whereas CPB1, AKR1C1, and CA2 showed lower expression levels (Fig.  4 C). Correspondingly, ROC curve analysis confirmed that all six hub genes exhibited strong diagnostic ability in discriminating gastric cancer samples from standard controls (Fig.  4 D), consistent with the previous results. To quantitatively evaluate the interactions between plasticizers and their potential targets, we performed comprehensive molecular docking simulations. The calculated binding affinity values for all plasticizer-target pairs are summarized in Supplementary Table 5 , providing a systematic assessment of interaction strengths. Notably, several plasticizer-target combinations demonstrated binding affinities comparable to those of known inhibitors. Specifically, BBP exhibited strong binding to AKR1C1 (-11.2 kcal/mol), approaching the affinity of the established inhibitor AKR1C1-IN-1 (-11.4 kcal/mol). Similarly, DEHP showed a significantly higher binding affinity for TDO2 (-8.4 kcal/mol) than the reference inhibitor LM10 (-7.2 kcal/mol). These comparisons provide crucial benchmarks for evaluating the biological relevance of the identified interactions. The binding conformations and the key amino acids involved in the binding are illustrated in Fig.  5 A-H. The key binding sites of CPB1 and BBP are ALA and LYS. The binding sites for AKR1C1 and BBP are LEU and TYR, respectively. The binding sites of GRM2 and BBP are ALA, ASN, and ARG. The binding sites of GRM2 and DBP are ARG, TYR, SER, and ARG. The binding sites of CA2 and DBP are LEU, VAL, HIS, VAL, THR, and ALA. The binding sites of CA2 and DEHP are LEU, VAL, HIS, GLN, PHE, THR, and ALA. The binding sites of MMP7 and DEHP are LEU, HIS, GLN, ALA, TYR, and PHE. The binding sites of TDO2 and DEHP are LEU, VAL, HIS, PHE, MET, ARG, TRP, GLY. Fig. 5 ( A ) Docking outcomes of CPB1 with BBP. ( B ) Docking outcomes of AKR1C1 with BBP. ( C ) Docking outcomes of GRM2 with BBP. ( D ) Docking outcomes of GRM2 with DBP. ( E ) Docking outcomes of CA2 with DBP. ( F ) Docking outcomes of CA2 with DEHP. ( G ) Docking outcomes of TDO2 with DEHP ( A ) Docking outcomes of CPB1 with BBP. ( B ) Docking outcomes of AKR1C1 with BBP. ( C ) Docking outcomes of GRM2 with BBP. ( D ) Docking outcomes of GRM2 with DBP. ( E ) Docking outcomes of CA2 with DBP. ( F ) Docking outcomes of CA2 with DEHP. ( G ) Docking outcomes of TDO2 with DEHP

Four gastric cancer transcriptome datasets ( GSE65801 , GSE79973 , GSE118916 , and GSE27342 ) were retrieved from the NCBI GEO database. A detailed summary of these datasets is provided in Supplementary Table 1 . These datasets were selected for their relevance to our research hypotheses, which aim to identify key molecular biomarkers and potential therapeutic targets in gastric cancer. Specifically, GSE65801 (64 samples) and GSE79973 (20 samples) were utilized as training cohorts to develop and validate predictive models, while GSE118916 (30 samples) and GSE27342 (160 samples) served as test cohorts to assess the generalizability of the findings. This validation-focused design prioritizes assessment of model generalizability across diverse cohorts. These datasets represent diverse gastric cancer patient populations and have been extensively used in previous studies to explore gene expression profiles and molecular mechanisms underlying gastric cancer development [ 16 – 19 ]. In summary, these datasets are sufficiently representative and reliable and can be used in this study to test the research hypotheses. To eliminate batch effects, surrogate variable analysis (SVA) was first applied to correct for potential confounding factors [ 20 ]. Subsequently, Data distributions were further calibrated using the ComBat algorithm, improving consistency across samples. Finally, expression distributions before and after correction were visualized using box plots and principal component analysis (PCA) to assess the effectiveness of batch correction, and the proportion of variance attributable to technical batch was precisely quantified using principal variance component analysis (PVCA). The complete workflow is shown in Fig. 1 . Fig. 1 Workflow schematic for the dataset analysis Workflow schematic for the dataset analysis We collected common plasticizers from PubMed and Google Scholar, and then obtained the SMILES structures of their representative compounds from PubChem. Component-specific targets in Homo sapiens were then assembled from the ChEMBL, SwissTargetPrediction, and STITCH databases. Targets obtained from all sources were consolidated and deduplicated to create a comprehensive set of plasticizer-related targets. Differentially expressed genes within the gastric cancer datasets were determined using the limma package, with significance criteria of adjusted P   1.25. A linear model was fitted using lmFit, and empirical Bayes moderation was applied with the eBayes function. The differential expression profiles were graphically explored using ggplot2 and heatmap. To identify the candidate genes for further analysis, the DEGs were cross-referenced with the plasticizer-related targets, and the resulting intersection was displayed in a Venn diagram. Using the clusterProfiler package, we conducted Gene Ontology (GO) functional annotation and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses of the potential target genes ( P < 0.05) [ 21 ]. The study was based on the Homo sapiens genome annotation package org.Hs.eg.db. GO analysis covered biological processes (BP), cellular components (CC), and molecular functions (MF), whereas KEGG analysis was applied to uncover the principal mechanisms by which plasticizers contribute to gastric cancer onset and progression. We established a comprehensive machine learning framework to identify gastric cancer biomarkers associated with plasticizer exposure systematically. Gene expression data were first normalized using Z-scores and partitioned into a training set and an independent test set. A total of 113 predictive models were constructed using ten representative algorithms (Supplementary Table 2 ), and their performance was compared using a multi-metric evaluation system that included the Receiver Operating Characteristic (ROC) curve (AUC), accuracy, F1-score, sensitivity, and specificity, with model robustness ensured through cross-validation. The core genes were the specific features (variables) retained by this best-performing model and the model’s built-in feature selection mechanism during training determined this final gene set. The R packages ggplot2 and PerformanceAnalytics were applied to visualize gene expression distributions and intergene correlations across sample groups. RNA-seq TPM data and corresponding clinical information of patients with stomach adenocarcinoma (STAD) were retrieved from the Cancer Genome Atlas (TCGA) database. To further validate the differential expression of the six key genes between gastric cancer and normal tissues, expression levels were compared using the Wilcoxon rank-sum test. Boxplots illustrating the expression distribution of each gene across different sample types were generated with the ggplot2 package. In addition, based on the clinical profiles of STAD patients, the pROC package was used to construct ROC curves and to assess the diagnostic performance of these key genes in distinguishing tumor tissues. Molecular docking was performed to examine interactions between plasticizer components and key target proteins. The three-dimensional structures of the target proteins were obtained from the Protein Data Bank (PDB; http://www.rcsb.org/pdb ). Protein preparation involved removing water molecules and ions in PyMOL, adding hydrogen atoms, assigning partial charges, and converting to PDBQT format using AutoDock Tools. Ligand structures in SDF format were retrieved from PubChem and energy-minimized using the Merck Molecular Force Field (MMFF) in Chem3D Pro before conversion to PDBQT format. Docking simulations were performed using AutoDock Vina (version 1.1.2). Binding affinities were evaluated based on the calculated free energy values (kcal/mol), with the following thresholds applied: strong binding ( ≤ − 7.0 kcal/mol), moderate binding (− 7.0 to − 5.0 kcal/mol), and weak binding ( ≥ − 5.0 kcal/mol). The resulting docking poses were analyzed and visualized using Discovery Studio 2019. Furthermore, by searching the MdeChemExpress website, we obtained the binding energies of the key genes with the inhibitors (Supplementary Table 5 ).

This study represents the first multi-omics integration of network toxicology, molecular docking, and machine learning to elucidate how plasticizers may contribute to gastric carcinogenesis. For the first time, we identify a novel mechanism by which plasticizers contribute to the onset of gastric cancer by modulating key targets, including CPB1, AKR1C1, GRM2, CA2, MMP7, and TDO2, as well as their associated signaling pathways. This finding provides essential theoretical support and a research foundation for understanding the carcinogenic mechanisms of environmental chemicals, biomarker discovery, and exposure risk management.

Plasticizers, widely used in consumer products, medical devices, and food packaging, can potentially migrate into the human body, posing significant long-term health risks. Gastric cancer, a malignancy with high global incidence and mortality, has increasingly been linked to environmental influences such as exposure to plasticizers. Yet, the mechanisms by which plasticizers influence gastric carcinogenesis remain inadequately explored. Therefore, the present study integrates network toxicology, molecular docking, and machine learning approaches to systematically investigate the potential impact of plasticizers on gastric cancer from a multi-omics perspective. The study identifies key targets and elucidates the possible mechanisms of action of BBP, DBP, and DEHP in gastric cancer. Unlike previous toxicology studies that focused on single compounds or isolated pathways, our research provides the first systematic integration of network toxicology, molecular docking, and multi-algorithm machine learning to identify potential gastric cancer–specific targets related to plasticizer exposure. These findings offer novel insights into the causal relationship between plasticizer exposure and gastric cancer and provide valuable theoretical support for the development of future risk assessment models and prevention strategies. Using machine learning, six key genes (CPB1, AKR1C1, GRM2, CA2, MMP7, and TDO2) were further identified and play significant roles in plasticizer-induced gastric cancer. Functional enrichment analysis indicated that these genes participate in critical signaling pathways, notably in the linoleic acid and arachidonic acid metabolism. These metabolic pathways provide a potential mechanistic bridge linking plasticizer-induced molecular disruptions to gastric carcinogenesis. Among them, the linoleic acid metabolism pathway has gained increasing attention due to its involvement in gastric cancer. As a precursor of various bioactive lipids, linoleic acid has been shown to promote gastric cancer cell invasion and peritoneal metastasis through activation of the ERK signaling pathway [ 22 ]. It may also contribute to poor prognosis in gastric cancer patients by downregulating miR-126 expression [ 23 ], suggesting that linoleic acid metabolism could play a pro-carcinogenic role in plasticizer-induced gastric cancer progression. Additionally, the arachidonic acid metabolism pathway, a key lipid mediator involved in inflammation and immune regulation, has been shown to have a significant pro-carcinogenic effect when dysregulated. Arachidonic acid metabolites promote tumor growth, angiogenesis, invasion, and metastasis [ 24 ]. In gastric cancer, this pathway enhances tumor metastasis by up-regulating cyclooxygenase-2 and promotes gastric cancer cell proliferation by activating the Notch1 signaling pathway [ 25 ]. These findings suggest that plasticizers may drive gastric carcinogenesis by disrupting the inflammatory microenvironment mediated by this pathway. In addition, our study also found a significant enrichment of the GABA receptor complex. This complex regulates neurotransmitter transmission and ion balance within cells, thus influencing cell proliferation and survival. Previous studies have indicated that GABA receptor expression is often up-regulated in cancer cells, and modulation of GABA receptor activity may regulate tumor growth and immune infiltration [ 26 ]. Our analysis newly implicates plasticizer exposure as a potential upstream disruptor of these critical pathways in gastric cancer. The differential expression analysis identified upregulation of GRM2, MMP7, and TDO2, and downregulation of CPB1, AKR1C1, and CA2 in gastric cancer tissues. It is noteworthy that the diagnostic value of these six genes was confirmed through ROC analysis, highlighting their potential as clinical biomarkers. A key novel finding of our study is the identification of GRM2, a metabotropic glutamate receptor, which primarily regulates neurotransmission and neuronal excitability [ 27 ]. Although GRM2 has been studied in the context of psychiatric disorders, its role in gastric cancer remains relatively unexplored. This study is the first to report its potential carcinogenic role in gastric cancer, suggesting that plasticizers may promote tumorigenesis by activating glutamate signaling. MMP7, a metalloproteinase, is deeply involved in both the initiation and advancement of gastric carcinoma. Previous investigations have shown that high expression of MMP7 is associated with tumor growth, drug resistance, and poor prognosis in gastric cancer [ 28 , 29 ]. As a significant carcinogenic factor, Helicobacter pylori infection can also induce increased gastrin secretion, which stimulates MMP7 release [ 30 ].Additionally, DEHP has been shown to mediate tissue remodeling in the male reproductive system via MMP7, suggesting that plasticizers may share common pathogenic pathways across various diseases. TDO2, a key enzyme in the tryptophan-kynurenine axis, is an independent prognostic factor for gastric cancer and is closely linked to immune responses [ 31 ]. While DEHP metabolites are known to up-regulate TDO2 in uterine leiomyoma cells [ 32 ], our study newly establishes the correlation between high TDO2 expression in gastric cancer and plasticizer exposure, suggesting that TDO2 may serve as a target for DEHP’s carcinogenic effects. CPB1, a novel biomarker associated with pancreatic and breast cancers [ 33 , 34 ], was found to be downregulated in gastric cancer and linked to BBP exposure. This association of CPB1 with plasticizer-induced gastric carcinogenesis is a novel observation. Previous studies have suggested that AKR1C1 is associated with increased endometriosis following plasticizer exposure [ 35 ]. In our study, we observed that AKR1C1 expression is reduced in gastric cancer, and this reduced expression may be closely associated with BBP exposure. Therefore, the potential role of AKR1C1 as a biomarker in gastric cancer warrants further investigation, particularly regarding the association between plasticizer exposure and gastric cancer development. Similarly, our study provides the first evidence linking the downregulation of CA2 in gastric cancer to plasticizer exposure. CA2, a zinc-metalloenzyme, catalyzes the reversible hydration of carbon dioxide and plays a key role in adjusting the pH of the tumor microenvironment [ 36 ]. CA2 is aberrantly expressed in several cancers [ 37 – 39 ]. Our findings suggest that plasticizers may contribute to tumor progression by altering the acidic microenvironment through CA2 suppression. Further correlation analysis revealed a positive correlation between CPB1 and CA2 expression, a negative correlation between MMP7 and AKR1C1 expression, and a positive correlation between TDO2, GRM2, and MMP7 expression. In hepatocellular carcinoma, previous studies have reported that TDO2 promotes liver cancer cell migration and invasion by up-regulating MMP7 [ 40 ]. These findings highlight interactions among different genes and provide clues to the potential molecular mechanisms underlying the link between plasticizers and gastric cancer development. To explore potential direct interactions between plasticizers and the identified key targets, we performed molecular docking analysis. The molecular docking results suggested specific interactions between plasticizers (BBP, DBP, and DEHP) and six core genes (CPB1, AKR1C1, GRM2, CA2, MMP7, and TDO2). The findings indicate that the binding sites of each plasticizer differ across genes, providing structural insights into how these genes might be involved in the functional regulation of gastric cancer cells. Specifically, the interaction between BBP and GRM2 could potentially disrupt the glutamate metabolism pathway, possibly altering gastric cancer cell proliferation. The glutamate metabolic pathway is a critical component of tumor cell energy metabolism. Therefore, the BBP-GRM2 interaction might represent one potential mechanism in regulating cancer cell proliferation and survival. Additionally, binding of BBP to CPB1 and AKR1C1 could hypothetically alter kinase activity, thereby impacting cell signaling pathways that depend on these enzymes. AKR1C1 is closely associated with kinase activity, whereas CPB1, as a potential biomarker, may regulate downstream signaling pathways upon interaction with plasticizers, contributing to gastric cancer progression. Interactions between plasticizers and these targets could alter the expression of related genes, disrupt normal cellular processes, and further promote the development of gastric cancer. Particularly noteworthy are several interactions that demonstrate binding affinities comparable to, or even stronger than, those of established inhibitors. Specifically, BBP showed strong binding to AKR1C1, approaching the affinity of the known inhibitor AKR1C1-IN-1. Similarly, DEHP exhibited significant binding to TDO2, with an affinity exceeding that of the reference inhibitor LM10. These quantitative comparisons provide strong evidence for the potential biological significance of these interactions, suggesting that plasticizers may compete with endogenous ligands or known inhibitors for binding to these key targets. Based on our research, we have proposed a coherent mechanistic hypothesis regarding the carcinogenic effects of plasticizers on gastric cancer. We believe that exposure to plasticizers simultaneously triggers oxidative stress (through lipid metabolism disorders), endocrine disorders (through interference with neurotransmitter and hormone signal transduction), and metabolic disturbances (through changes in enzyme activity and pathway regulation). These molecular damages interact with each other, leading to dysregulation of key pathways—including linoleic acid and arachidonic acid metabolism and neuroendocrine signal transduction —ultimately resulting in tumor proliferation, invasion, and immune evasion. This pathogenic cascade reaction is driven by the dysregulation of six key genes: GRM2, MMP7, TDO2, CPB1, AKR1C1, and CA2. Although this study provides novel insights into the potential mechanisms linking plasticizer exposure to gastric cancer development, several limitations should be acknowledged. First, the findings were primarily derived from bioinformatics analyses and network toxicology predictions, which may introduce inherent biases stemming from database selection, data integration, and algorithmic assumptions. Second, the molecular docking results represent theoretical predictions that require experimental validation; they should be considered as hypothesis-generating rather than conclusive evidence of mechanistic interactions. Third, no in vivo or in vitro experiments were performed to validate the computational results. Therefore, future studies should conduct wet-lab experiments, including qPCR, Western blot, and gene knockdown assays, to confirm the identified biomarkers and mechanisms. And, to validate the stability of the binding conformation (for example, through the root-mean-square deviation (RMSD) values), further exploration of molecular dynamics simulations is still needed [ 41 , 42 ]. Moreover, the dose–response relationship between plasticizer exposure and gastric cancer risk remains to be quantitatively characterized. Integrating dose–response modeling with clinical and epidemiological data would enhance the translational value of these findings.

Plasticizers, a group of chemical additives, are widely incorporated into plastic products to enhance their flexibility and durability. Common examples of plasticizers include benzylbutyl phthalate (BBP), dibutyl phthalate (DBP), and di-(2-ethylhexyl) phthalate (DEHP) [ 1 ]. Due to their favorable physicochemical properties, these plasticizers are pervasive in life. Nevertheless, their potential risks have become increasingly evident as research on their health effects advances [ 2 ]. Studies have shown that plasticizers’ lipophilic nature and environmental stability contribute to their significant bioaccumulation in the human body, thereby increasing long-term health risks [ 3 , 4 ]. A growing body of research suggests that plasticizers may pose a potential carcinogenic risk. For instance, studies have shown that plasticizers can enhance the proliferation and invasive capacity of liver cancer cells through the IRE1α-XBP1s axis [ 5 ] and that elevated exposure to plasticizers has been linked to a higher incidence of colorectal cancer [ 6 ]. In ovarian cancer, plasticizers contribute to tumor progression by activating the PI3K/AKT pathway [ 7 ]. Additionally, the DNA-damaging effects of plasticizers and their potential disruption of cellular metabolism have become important areas of focus in cancer research [ 8 , 9 ]. Thus, elucidating the association between plasticizers and various types of cancer is critical for assessing public health risks. Gastric cancer ranks as the fifth most frequently diagnosed malignancy and the fourth principal contributor to cancer mortality. Its development is influenced by various factors, including genetic susceptibility, Helicobacter pylori infection, dietary habits, and environmental exposures [ 10 ]. In addition to these established risks, emerging evidence suggests that plasticizer exposure could be a significant factor in both the initiation and development of gastric cancer. Evidence indicates that co-exposure to Helicobacter pylori and plasticizers may compromise the integrity of the gastric mucosa, thereby increasing the risk of gastric cancer [ 11 ]. Moreover, studies have shown that low-dose exposure to DEHP exacerbates the epithelial-mesenchymal transition in gastric cancer cells via the mTOR and Smad2 signaling pathways, thereby promoting tumor progression [ 12 ]. Despite some progress in research, however, the specific molecular targets, signaling pathways, and potential mechanisms by which plasticizers contribute to gastric cancer remain poorly understood, warranting further in-depth investigation. To date, no study has systematically elucidated these mechanisms using an integrative analytical framework that combines target prediction, pathway enrichment, and molecular docking. Network toxicology, as an innovative interdisciplinary approach, combines bioinformatics, systems biology, and cheminformatics. It provides a robust framework for understanding how chemicals disrupt biological molecular networks, impair cellular homeostasis, and ultimately lead to disease. Integrating and analyzing extensive biological and toxicological data enables the systematic identification of potential targets and the inference of chemical toxicity mechanisms [ 13 ]. Recent network toxicology analyses of plasticizers have deepened our understanding of their carcinogenic potential. For instance, studies linking DEHP to breast and prostate cancers have revealed disruptions in critical pathways such as tumor signaling and focal adhesion, as well as regulatory effects on genes involved in post-translational modification and extracellular matrix organization [ 14 , 15 ]. These pioneering investigations demonstrate the feasibility of network toxicology in elucidating chemical–cancer interactions. Despite these advances, current research has primarily examined individual plasticizers, employed relatively narrow bioinformatics methods, and rarely incorporated multi-omics integration. Notably, the network toxicology approach has yet to be systematically applied to uncover the molecular mechanisms of plasticizer-induced gastric carcinogenesis. Molecular docking can offer a complementary strategy for simulating atomic-level interactions between plasticizers and specific protein targets—such as receptors and enzymes—providing a structural basis for understanding their potential roles in cancer pathogenesis. In this context, the present study integrates network toxicology with molecular docking to examine the relationships between common plasticizers and key protein targets relevant to gastric carcinoma. By combining data from multiple public databases with bioinformatics analysis and machine learning algorithms, potential target genes regulated by plasticizers are identified and subsequently validated through molecular docking to assess potential binding interactions with core proteins. This study seeks to elucidate the molecular mechanisms of plasticizer-induced gastric cancer. It also provides a theoretical foundation for environmental exposure risk assessment and the design of appropriate interventions.

Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1 Supplementary Material 2 Supplementary Material 2 Supplementary Material 3 Supplementary Material 3 Supplementary Material 4 Supplementary Material 4 Supplementary Material 5 Supplementary Material 5 Supplementary Material 6 Supplementary Material 6