TELO2 mediates parabens-induced breast carcinogenesis: a comprehensive network analysis Authors: Jing Ren, Xiaofen Li, Bin Dong, Zhulin Bu, Yuhui Wu, Yuting Li, Lin Yang, Huaixi Xing, Yuting Dai, Shuosheng Zhang, Xianglong Meng Published: 2025 ABSTRACT AbstractBackgroundParabens (PBs) are associated with an increased risk of breast cancer, yet their underlying molecular mechanisms remain poorly understood. This study aimed to comprehensively elucidate the targets and mechanisms of PBs in breast cancer by integrating network toxicology, bioinformatics, Mendelian randomization (MR), molecular docking, and other complementary methodologies.ResultsNetwork toxicology analysis identified 2,851 potential PB targets, with 172 significantly linked to breast cancer. Pathway enrichment revealed that PBs predominantly influence the Phosphatidylinositol 3-kinase-Akt (PI3K-Akt) signaling pathway and the cell cycle pathway. Two-sample MR identified TELO2 as a significant risk factor for both malignant and benign breast cancer (malignant: IVW OR = 1.06, 95% CI: 1.001–1.126, p = 0.047; benign: IVW OR = 1.13, 95% CI: 1.009–1.270, p = 0.034). Bioinformatics analysis demonstrated that TELO2 expression was significantly elevated in breast cancer tissues (p < 0.05) and exhibited high diagnostic accuracy (AUC: 0.803 in TCGA and 0.876 in GSE20685). Furthermore, mediation analysis revealed that modulating natural killer T (NKT) cells significantly mediated the TELO2-breast cancer link, with a mediation proportion of 20.46%. Molecular docking confirmed stable binding interactions between PBs and the TELO2 protein. Moreover, an mRNA–microRNA (miRNA)–long non-coding RNA (lncRNA) regulatory network centered on TELO2 identified 19 miRNAs and 189 lncRNAs as potential regulators of its expression.ConclusionsOur integrative findings suggest that parabens may exert deleterious effects in the context of breast cancer by specifically targeting the TELO2 gene and its associated regulatory networks and pathways. These findings not only advance our understanding of the environmental drivers of BC but also pave the way for future research aimed at mitigating the disease’s health burden through targeted interventions against harmful environmental exposures.Graphical Abstract Supplementary InformationThe online version contains supplementary material available at 10.1186/s40360-025-01072-1.Keywords: Parabens, Breast cancer, Network toxicology, Mendelian randomization, Bioinformatics Introduction Parabens (PBs), such as methylparaben (MP), ethylparaben (EP), propylparaben (PP), and butylparaben (BP), are widely used preservatives in cosmetics, food, and pharmaceuticals [1]. Scientific advancements have raised concerns regarding their ecological and health impacts [2]. Regulatory agencies in the U.S., EU, China, and Japan have established strict limits on PBs to mitigate risks [3–6]. Used for decades in foods like beverages, grains, and seasonings [7], PBs are approved as food additives by the FDA and have been detected globally in food and environmental samples [8–10]. Dermal absorption from personal care products is a major exposure route [11], and PBs are also present in pharmaceuticals [12]. Their prevalence in consumer products leads to environmental detection in water, air, and soil in regions such as Brazil and Spain [13–15]. Human exposure occurs via respiratory, dermal, and dietary pathways, making it critical to identify sources, quantify concentrations, and assess associated health risks. Upon entering the human body, PBs are absorbed into the bloodstream and metabolized by nonspecific esterases via enzymatic hydrolysis, followed by conjugation with sulfate or glucuronic acid [16]. These metabolites are primarily excreted in urine [17]. PBs have also been detected in blood, adipose, and placental tissues [18–20]. Evidence suggests that PB exposure adversely affects male reproductive parameters, including sperm concentration, count, and motility [21]. Acting as endocrine disruptors, PBs bind to estrogen receptors, mimicking or antagonizing estradiol [22]. BP shows the highest estrogenic activity—about 1/100,000th that of estradiol—though typical human exposure remains below effect levels [23]. National Health and Nutrition Examination Survey (NHANES) analyses indicated an inverse correlation between urinary paraben (PB) levels and diabetes (OR = 0.60–0.84), notwithstanding the possibility of reverse causality [24]. In a Spanish cohort, serum methylparaben was associated with elevated prostate cancer risk (HR = 1.60, 95% CI: 1.16–2.20) [25]. A Chinese case–control study linked urinary propylparaben to thyroid cancer (OR = 0.24, 95% CI: 0.18–0.31) [26], and propyl p-hydroxybenzoate was significantly higher in lung cancer patients (OR = 2.22, p < 0.001) [27]. MP produces photoproducts under sunlight exposure, which exhibit DNA-damaging activity in vitro, particularly in the presence of divalent copper ions, with reactive oxygen species (ROS) involved [28]. Parabens’ metabolites can also induce oxidative stress via ROS to cause DNA damage; their exposure correlates with male reproductive health [21], and urinary parabens concentrations are significantly associated with 8-hydroxy-2’-deoxyguanosine (8-OHdG) levels, supporting their role in oxidative stress [24]. Breast cancer (BC) is a malignant tumor with high metastatic potential that severely impacts patient survival and quality of life [29]. Persistent residues of PBs such as MP in breast tissue pose potential risks; in vitro studies demonstrate that PBs, particularly in combination, enhance MCF-7 breast cancer cell proliferation in a dose-dependent manner, with significant increases observed at concentrations ≥ 1 µM [30]. Clinically, higher PB levels are found in ER+/PR + breast tumor tissues compared to healthy tissues (OR = 98.34, P = 0.027) [31]. Long-chain esters like BP and PP weakly bind estrogen receptors, activating estrogen-responsive genes and promoting hormone-dependent proliferation [31]. Recent evidence also indicates that PBs inhibit 17β-hydroxysteroid dehydrogenase (17β-HSD1/2), elevating local estrogen levels and potentially advancing ER + breast cancer [32]. While numerous epidemiological and in vitro studies indicate PBs may adversely affect breast cancer, the exact mechanisms remain unclear. Advancements in science and technology have enabled the development of diverse methods to investigate the relationships and mechanisms linking diseases and environmental pollutants. Bioinformatics applies computational and statistical tools to analyze biological data and uncover molecular mechanisms. Network toxicology constructs models interconnecting chemicals, targets, and toxic outcomes, supporting mechanistic and risk prediction for environmental toxicants. Mendelian randomization (MR) uses genetic variants as instrumental variables to infer causal exposure-disease relationships while reducing confounding bias. Additionally, long-term epidemiological data help assess associations among exposures, health outcomes, and risk factors. Collectively, these approaches provide an integrative framework for studying the effects and mechanisms of PBs on breast cancer. This study utilized an integrated approach incorporating network toxicology, Mendelian randomization, molecular docking, molecular dynamics simulations, and bioinformatics to systematically evaluate the impact of PBs on breast cancer and elucidate the underlying molecular mechanisms. For the first time, this comprehensive multi-methodological strategy identified telomere maintenance protein 2 (TELO2) as a key mediator gene in PB-induced breast carcinogenesis. These findings address a critical knowledge gap regarding the role of TELO2 in environmental pollutant-driven breast cancer pathogenesis. The study provides novel molecular targets for the prevention and treatment of environment-associated cancers and offers new insights for a comprehensive assessment of breast cancer risk related to environmental PB exposure, thereby underscoring the innovative and scientific contributions of this work. Materials and methods This study employed network toxicology to explore the relationship between primary PBs targets and breast cancer, followed by Mendelian randomization to assess the causal links between potential targets, exposure, and outcomes in BC. Bioinformatics tools were used to analyze target expression profiles in breast cancer samples. Evaluative and Exploratory Study on the Impact of PBs on Breast Cancer Using NHANES Data. Figure 1 illustrates the technical workflow, and Table 1 summarizes the key databases and software used. Technology roadmap. This study integrated multi-methodologies to explore telomere maintenance protein 2 (TELO2) as the mediator in parabens (PBs)-induced breast cancer (BC). Step 1: Evaluate toxicity of four PBs [methylparaben (MP), ethylparaben (EP), propylparaben (PP), butylparaben (BP)] via ProTox-3.0/ADMETlab 3.0 and collect their targets from six databases. Step 2: Retrieve BC targets from GeneCards/TCGA/GEO and perform differential expression analysis with DESeq2/limma. Step 3: Identify PBs-BC overlapping targets, construct PPI network, and conduct GO/KEGG enrichment. Step 4: Verify causal relationships via two-sample MR and explore NKT cell mediation. Step 5–7: Perform TELO2 bioinformatics analysis (ROC, survival, immune infiltration, scRNA-seq), verify PBs-TELO2 binding (docking/MD), and construct mRNA-miRNA-lncRNA network. Step 8: Analyze PBs-BC association via NHANES. All analyses used RStudio and Cytoscape Table 1List of databases and softwareDatabasesDatabase linksPublic Chemical Database, Pubchem https://pubchem.ncbi.nlm.nih.gov/ Swiss Target Prediction http://www.swisstargetprediction.ch/ Universal Protein Resource, UniProt https://www.uniprot.org/ GeneCards-The Human Gene Database, GeneCards https://www.genecards.org/ Protein Data Bank, PDB http://www.rcsb.org/ Xiantao Academic https://www.xiantaozi.com/ ProTox 3.0 - Prediction Of Toxicity Of Chemicals, ProTox 3.0 https://tox.charite.de/protox3/index.php?site=home ADMETlab, ADMETlab 3.0 https://admetlab3.scbdd.com/ Super-PRED https://prediction.charite.de/ Super-Enhancer Archive, SEA https://sea.bkslab.org/ Bioinformatics Analysis Tool for Molecular mechANism of Traditional Chinese Medicine, BATMAN-TCM http://bionet.ncpsb.org.cn/batman-tcm/index.php/Home/document/index Comparative Toxicogenomics Database, CTD https://ctdbase.org/ Search Tool for Interactions of Chemicals, STITCH http://stitch.embl.de/ FinnGen https://www.finngen.fi/en Genotype-Tissue Expression Project, GTEx https://gtexportal.org/home/ Genome-Wide Association Studies Catalog, GWAS catalog https://www.ebi.ac.uk/gwas/ Cell Identification By Estimating Relative Subsets Of RNA Transcripts (x), CIBERSORTx https://cibersortx.stanford.edu/ Cell Marker Database 2.0, CellMarker 2.0 http://117.50.127.228/CellMarker/index.html miRNA Target Database, miRDB https://mirdb.org/ DIANA-MicroT-CDS (microRNA Target Prediction), DIANA-MicroT https://dianalab.e-ce.uth.gr/microt_webserver/#/interactions miRWalk - miRNA Target Prediction, miRWalk http://129.206.7.150/ The Encyclopedia of RNA Interactomes, ENCORI/starBase https://rnasysu.com/encori/index.php Search Tool for the Retrieval of Interacting Genes/Proteins, STRING https://string-db.org/ Human Protein Atlas, HPA https://www.proteinatlas.org/ National Health and Nutrition Examination Survey, NHANES https://www.cdc.gov/nchs/nhanes/ Cavity-detection guided Blind Docking 2, CB-dock2 https://cadd.labshare.cn/cb-dock2/php/index.php RCSB Protein Data Bank, RCSB https://www.rcsb.org/ The Cancer Genome Atlas Program, TCGA https://www.cancer.gov/ccg/research/genome-sequencing/tcga CytoscapeCytoscape Consortium, 3.9.0RStudioPublic Benefit Corporation, 4.3.3PymolSchrödinger, Limited Liability Company, 2.6 List of databases and software Toxicity identification and target collection of PBs The ProTox-3.0 (https://tox.charite.de/protox3/index.php?site=home) and ADMETlab databases (https://admetlab3.scbdd.com/) were utilized to evaluate the environmental toxicity of MP, EP, PP, and BP [33, 34]. To facilitate further analysis, the PubChem database (https://pubchem.ncbi.nlm.nih.gov/) was queried to obtain SMILES codes, PubChem CID numbers, and SDF format files for these four compounds. Relevant keywords were employed to search the CTD (https://ctdbase.org/), STITCH (http://stitch.embl.de/), and Super-PRED databases (https://prediction.charite.de/) for information related to the compounds [35]. Additionally, the SwissTargetPrediction and SEA databases (https://sea.bkslab.org/) were accessed to identify potential targets linked to these compounds based on their SMILES code [36]. The Batman database (http://bionet.ncpsb.org.cn/batman-tcm/index.php/Home/document/index) was also queried using the PubChem CID numbers to retrieve corresponding targets [37]. All database searches were limited to the species “Homo sapiens”, with specific selection criteria: the SwissTargetPrediction database required a probability > 0.01, the Super-PRED database required a probability of 50% or higher, and the Batman database required a correlation score >10. The UniProt IDs of the targets identified in the Super-PRED and SwissTargetPrediction databases (http://www.swisstargetprediction.ch/) were converted into gene symbols using the UniProt database (https://www.uniprot.org/). Ultimately, the target prediction results from six databases were consolidated to create a unique target dataset for PBs. A comprehensive target set was then formed by merging the target datasets of the four parabens, and a Venn diagram was generated using Xiantao Academic (https://www.xiantaozi.com/) to provide essential data support for future research. Breast cancer targets collection The GeneCards database (https://www.genecards.org/) was queried for ‘Breast Cancer’ with a filter for ‘Homo sapiens’, selecting protein-coding genes with a relevance score > 2. Data from the TCGA database, comprising 1,113 breast cancer cases and 113 controls, were curated (https://www.cancer.gov/ccg/research/genome-sequencing/tcga). Differential gene expression analysis was conducted using the DESeq2 package (Version 1.36.0) in RStudio (Version 4.3.3) [38]. The GEO database was also queried, integrating the GSE20685 dataset with 104 breast cancer cases and 17 controls (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE20685), analyzed using the limma package (Version 3.58.1) in RStudio [39]. DEGs were identified in both datasets with the criteria of an adjusted p-value < 0.05 and |logFC| ≥ 1. A volcano plot was created using ggplot2 (Version 3.4.4) in RStudio, and overlapping breast cancer targets were visualized with a Venn diagram via the Xiantao Academic platform. Target and enrichment analysis in breast cancer treated with PBS To explore potential paraben targets in breast cancer, we first examined the overlap between parabens and breast cancer-related targets. A protein–protein interaction network was constructed using the STRING database, specifying ‘Homo sapiens’ and a confidence threshold of 0.15 to identify the edges, nodes, and degrees of target genes. Topological parameters were calculated with Cytoscape software (Version 3.9.0), using degree value as a key metric, where darker node colors indicated higher degree values and stronger interactions. Cytoscape’s network property calculations helped assess the significance of connections. We conducted Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses on the intersecting targets using the clusterProfiler package (Version 4.4.4) in Rstudio software to explore the biological functions and pathways related to the risk of breast cancer associated with benzoate exposure, with the subjects being humans. This study aimed to assess the impact of parabens on potential breast cancer-related targets, categorized into three groups. The enrichment analysis results for biological process (BioP), cellular component (CC), and molecular function (MF) were visually represented using bar graphs created with the ggplot2 package in RStudio. Statistical significance was assessed with a threshold of p.adjust < 0.05. Mendelian randomization analysis This study used a two-sample Mendelian randomization (MR) approach with the TwoSampleMR package (Version 0.5.7) in RStudio to examine the causal association between expression levels of potential breast cancer targets and benign or malignant outcomes in the Finnish population [40]. Instrumental variables (IVs) were selected from GTEx (https://gtexportal.org/home/), focusing on expression quantitative trait loci (eQTLs) significantly linked to breast tissue targets [41]. Outcome data came from FinnGen (https://www.finngen.fi/en), including 3,574 breast cancer cases and 207,296 controls. All data were publicly available, ensuring ethical compliance. Single Nucleotide Polymorphisms (SNPs) strongly associated (p < 1e− 5) with gene expression were identified as IVs. To avoid weak instrument bias, SNPs with F-statistics < 10 were excluded. Linkage disequilibrium was minimized using r2 < 0.1 within 100 kb windows. The inverse-variance weighted (IVW) method was used for MR analysis, with significance determined by Bonferroni correction. Horizontal pleiotropy was assessed via MR-Egger regression (intercept p > 0.05 indicating no bias), and heterogeneity was evaluated using Cochran’s Q test (p > 0.05 indicating homogeneity). Genes with OR > 1 and 95% CI were considered risk factors. Multiple testing was corrected using FDR. Robustness was confirmed by leave-one-out analysis and visualized with forest plots. The previously identified genes were confirmed as targets for assessing PBs’ effects on breast cancer. We investigated the causal relationships between these genes and benign breast cancer risk, focusing on immune cells. Summary statistics for 731 immune cell traits were obtained from the GWAS Catalog (https://www.ebi.ac.uk/gwas/) (GCST90001391–GCST90002121) [42]. The study included over 22 million SNP genotypes from 3,757 individuals of European ancestry, with adjustments for sex and age. For instrument selection, we used a genome-wide significance threshold of p < 5e− 8 and applied FDR correction across all immune cell traits. We selected SNPs associated with 731 kind of immune cells at p < 5e− 8, excluded those in linkage disequilibrium, and retained strong instruments (r² = 0.1, Kb = 10,000 kb, F-statistic > 10). A two-step mediating MR analysis was performed: first, IVW-based MR assessed the association between immune cells and benign breast cancer risk (FDR-corrected p < 0.05); second, MR evaluated genes and immune cells with causal links, selecting mediators at FDR-corrected p < 0.05. We estimated several effects: total effect of immune cells on breast cancer (β_all), gene-to-immune-cell effect (β_1), immune-cell-to-cancer effect (β_2), mediated effect (β_12 = β_1 × β_2), direct effect (β_dir = β_all − β_12), and mediation proportion (β_12_p = β_12 / β_all). Bioinformatic analysis of target genes Differential expression of the target gene was assessed in BC samples from TCGA and GSE20685 datasets. Using ggplot2 in R, we visualized expression patterns between diseased and normal tissues and constructed Receiver Operating Characteristic (ROC) curves to quantify the diagnostic performance of gene expression levels. The pROC package in R [43] was used to compute the area under the curve (AUC), where a value above 0.7 signifies diagnostic significance. Using the survival package in R, Kaplan–Meier analysis evaluated the overall survival (OS) of breast cancer patients at clinical stages N0 and N2, with groups defined by the lowest p-value. The Cox proportional hazards model was used to calculate hazard ratios (HRs) along with their 95% confidence intervals. Furthermore, the clinical relevance of the gene was evaluated by analyzing the T stage, N stage, overall stage, age, dissection type, breast cancer subtype, and gender information from the TCGA database, with group comparison plots generated using the ggplot2 package in RStudio. Pathological section images of the gene in both the disease and normal groups were also compared using data from the HPA database (https://www.proteinatlas.org/). Using the TCGA breast cancer database, samples were stratified by target gene expression levels. Immune cell infiltration within the tumor microenvironment was then evaluated by Single-sample Gene Set Enrichment Analysis (ssGSEA), which calculates enrichment scores of immune cell marker genes [44]. The R package GSVA was employed to analyze and compare the expression profiles of 24 different immune cell types between the high and low expression cohorts [45]. The CIBERSORT deconvolution algorithm was applied to quantitatively assess the relative proportions of immune cell subtypes in heterogeneous tissue samples using gene expression data, based on 22 immune cell markers from the CIBERSORTx database (https://cibersortx.stanford.edu/) [46]. The relationships between target genes and immune cells, identified via ssGSEA and CIBERSORT, were analyzed, and correlation plots were generated using the ggplot2 package in RStudio. We performed single-gene differential expression analysis using DESeq2 on TCGA breast cancer samples. Gene set enrichment analysis (GSEA) was conducted with clusterProfiler, employing the “c2.cp.all.v2022.1.Hs.symbols” canonical pathways set from msigdbr [47]. Significant results were determined using a p.adjust threshold of less than 0.05. Finally, signaling pathways related to the PBs-BC effect were examined, and a mountain range map was created to visualize the findings effectively. To complement the bulk tissue analyses and to investigate the expression of TELO2 at a cellular resolution within the tumor microenvironment, we utilized the single-cell dataset GSE176078 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE176078). From the 26 available primary tumor samples, seven were randomly selected for downstream analysis. This approach provides a computationally efficient yet statistically powerful means to profile gene expression across diverse cell types, as the data from thousands of individual cells per sample adequately captures the cellular heterogeneity. Using RStudio software, the R packages Seurat, scDbIFinder, and Harmony were employed to preprocess the data by filtering out low-expression genes, removing doublets, and accounting for cell cycle effects. UMAP was used for dimensionality reduction and clustering, and cell annotation was conducted with the CellMarker2.0 database (http://117.50.127.228/CellMarker/index.html) to assess the target gene’s expression profile in immune cell populations [48]. Molecular docking and dynamics analysis A molecular docking approach was employed to predict binding free energy and evaluate interaction forces, confirming PBs’ binding capability to the target protein. The 3D structures of MP, EP, PP, and BP were retrieved from the PubChem database, while the target protein structures were obtained from the RCSB Protein Data Bank (http://www.rcsb.org/). PyMOL (Version 2.6) was used to remove the native ligand from the target protein, and molecular docking simulations were conducted online using the CB-DOCK2 platform (https://cadd.labshare.cn/cb-dock2/php/index.php) [49]. Molecular dynamics simulations of the MP and target gene were conducted using Amber24 with the ff19SB force field and OPC water model [50]. The complex was placed in a cubic water box with a cutoff distance of 1.0 nm. Electrostatic and van der Waals interactions were computed using a 2 fs time step, employing the PME method for long-range electrostatic corrections. The system was kept at 300 K and 1 bar. After the initial setup, energy minimization was performed, followed by 200 ps of NVE equilibrium dynamics and 100 ps of NPT equilibrium dynamics. The temperature was regulated via the V-rescale method, and the pressure was controlled using the Parrinello–Rahman approach. A 100 ns molecular dynamics simulation was conducted, during which the RMSD, RMSF, radius of gyration, and free energy landscape (FEL) were evaluated. mRNA–miRNA–IncRNA interaction network and gene interaction analysis Target genes were queried in the miRDB, DIANA-MicroT, miRWalk, and ENCORI/starBase databases [51, 52]. The intersection of the predicted miRNAs from these databases was considered as the group of candidate target genes. The shared miRNAs identified in the starBase database were used to predict the associated lncRNAs. An mRNA–miRNA–lncRNA network for the target genes was constructed using Cytoscape software. Additionally, the GeneMANIA database was used to analyze target genes for related mRNAs, gene interactions, pathways, and co-expression networks [53]. Cross sectional analysis via the NHANES database We analyzed data from 40,790 participants across four NHANES cycles (2005–2012). After excluding individuals with missing outcome variables, aged < 20 years, or lacking BMI or paraben measurements, 6,847 participants were included. The study received ethical approval, and informed consent was obtained. Covariates, including age, gender, race, income, education, smoking, drinking, BMI, diabetes, and hypertension, were selected based on literature. Group differences were evaluated using t-tests, Kruskal-Wallis tests, or chi-square tests. Multivariable regression employed three hierarchical models: Model 1 (exposure variables only), Model 2 (adding significant univariate covariates), and Model 3 (all significant exposure variables and potential covariates). All analyses applied sample weights to account for complex survey design. Results Evaluation of the toxicity of PBs and identification of potential toxicological targets The ProTox3.0 and ADMETab3.0 databases were used to evaluate the physicochemical properties and potential toxicity of MP, EP, PP, and BP. A positive correlation was observed between the lipophilicity and plasma protein binding capacity (PPB) of these compounds and their molecular weight (Fig. 2A). Paraben derivatives showed carcinogenic potential, indicated by their bioconcentration factor (BCF), inhibitory growth concentration at 50% (IGC50), lethal concentration at 50% for Danio rerio (LC50DM), lethal concentration at 50% for Daphnia magna (LC50FM), and overall carcinogenicity assessments (Fig. 2B). These results underscore the significant environmental toxicity and provide a basis for further research on the toxicological effects of these compounds on human health. The toxicity of PBs and the targets PPI and enrichment in breast cancer. (A) The physicochemical properties of PBs. (B) The environmental toxicity of PBs. (C) Venn diagram of PBs targets. Volcano plots displaying differentially expressed genes (DEGs) in both the TCGA and GSE20685 datasets. (F) Venn diagram of breast cancer targets. (G) Venn diagram of PBs and breast cancer intersection targets. (H) PPI network of PBs-BC targets. (I) GO and KEGG enrichment analyses To further investigate the potential targets of PBs, a comprehensive search was conducted across multiple databases, including CTD, STITCH, Super-PRED, SwissTargetPrediction, SEA, and Batman. This analysis identified 1134 targets for MP and 992 targets for EP. Additionally, 679 targets were found for PP and 983 targets for BP (Supplementary Table 1). Collectively, when combining the targets of these four compounds, a total of 2851 targets for PBs were obtained (Fig. 2C). Statistics on breast cancer targets To screen breast cancer targets, 8,375 candidates were retrieved from the GeneCards database (Supplementary Table 2). Differential expression analysis of 1,226 breast cancer samples in TCGA identified 5,077 differentially expressed genes (Fig. 2D). Similarly, analyzing 121 samples from GSE20685 in GEO identified 3,189 genes (Fig. 2E). Intersecting these results (Fig. 2F) revealed 832 common targets, which were then cross-referenced with PB-related targets, identifying 172 PB-associated breast cancer targets (Fig. 2G). PPI analysis and enrichment analysis of potential breast cancer targets affected by PBs The 172 PBs-treated breast cancer targets were analyzed using the STRING database (https://string-db.org/), excluding isolated targets. Cytoscape 3.9.0 was used to construct a PPI network with 172 nodes and 3771 edges, ranked by connectivity degree, where darker colors and larger circles denote stronger interactions (Fig. 2H), and the other topological analyses are provided in the Supplementary Table 3. GO and KEGG enrichment analyses were conducted with Rstudio, identifying 702 biological processes, 23 molecular functions, 23 cellular components, and 3 KEGG pathways (Fig. 2I). The GO analysis highlighted responses to steroid hormones, oxidative stress, hormone binding, regulation of cyclin-dependent protein serine/threonine kinase, and ligand-activated transcription factor activity. KEGG analysis showed enrichment in the PI3K-Akt signaling pathway, cell cycle regulation, and PPAR signaling pathways. Results of Mendelian randomization analysis A Mendelian randomization (MR) analysis was performed to assess the causal impact of PBs on breast cancer by examining the relationships between the identified targets and breast cancer outcomes. The 172 potential targets were designated as the exposure variables, while benign and malignant breast cancer cases from the Finnish database served as the outcome variables. Following the exclusion of targets lacking significant SNPs, 60 targets with significant SNPs were selected for further evaluation. After excluding confounding variables and SNPs in linkage disequilibrium, 47 targets remained for subsequent analysis. Subsequently, genes with more than three SNPs were screened, leading to the identification of 11 gene targets. The MR analysis using the IVW method yielded results with p-values below 0.05. In the context of malignant breast cancer outcomes, CDC7 and TELO2 were found to be significantly associated with breast cancer (Supplementary Table 4). Additionally, LARP6 and TELO2 demonstrated a significant causal relationship with breast cancer outcomes in benign tumor cases (Supplementary Table 5). Scatter plots (Fig. 3A-B) visually illustrate that as the effect of TELO2-associated SNPs on breast cancer increased, their influence on the outcome also grew stronger. These findings confirm the positive correlation between potential PB targets and breast cancer outcomes. Furthermore, leave-one-out sensitivity analysis (Fig. 3C-D) revealed a robust causal relationship between TELO2 and breast cancer, as the removal of any IV did not introduce significant bias or alter the effect size. The results of the MR analysis were depicted as forest plots (Fig. 3E), illustrating a significant association between TELO2 and outcomes in both malignant breast cancer (IVW OR = 1.06, 95% CI: 1.001–1.126, p = 0.047) and benign breast cancer (IVW OR = 1.13, 95% CI: 1.009–1.270, p = 0.034). These findings highlight TELO2’s potential role in promoting breast cancer development. Furthermore, Cochrane’s Q test was performed to assess heterogeneity among instrumental variables (IVs) for TELO2, and the results revealed no significant heterogeneity, with both Q-P values exceeding 0.05, indicating consistent effects of the included 4 SNPs for TELO2 on the outcome. Additionally, the p-values from the horizontal pleiotropy tests exceeded 0.05, effectively excluding the possibility of confounding distortions in the study results. Collectively, these genetic data provide compelling evidence for a causal relationship between TELO2 and the progression of breast cancer. Mendelian randomization results for TELO2 and BC. (A) Scatter plot illustrating Mendelian randomization outcomes for benign breast neoplasm. (B) Scatter plot illustrating Mendelian randomization outcomes for malignant breast cancer neoplasm. (C) Leave-one-out sensitivity for TELO2 on benign neoplasm of BC. (D) Leave-one-out sensitivity for TELO2 on malignant neoplasm of BC. (E) Forest plot depicting Mendelian randomization results for the effect of TELO2 on breast cancer A mediation MR analysis was conducted on 731 immune cell types to explore TELO2’s role in breast cancer etiology. After excluding cells with linkage disequilibrium and weak instrumental variables, 612 immune cells were analyzed. Benign breast cancer was the outcome variable, and the 612 immune cells were the exposure variables. The IVW method identified causal relationships in 25 immune cell types with p-values < 0.05 (Supplementary Table 6). Heterogeneity and horizontal pleiotropy analyses confirmed the stability of these results. Further MR analysis revealed causal links between TELO2 and four immune cell types: activated natural killer T cells (NKT AC), natural killer T cells (NKT cells), natural killer T lymphocytes (NKT lymphocytes), and CD20-positive IgD + CD38dim B lymphocyte subsets, all with ORs > 1 and p-values < 0.05 (Supplementary Table 7). Sensitivity analyses validated these findings, showing no horizontal pleiotropy or heterogeneity. A two-step mediation analysis indicated a positive association between TELO2 and breast cancer (OR = 1.13, β_all = 0.124). NKT immune cells exhibited the most significant mediating effect, being positively linked to benign breast cancer outcomes (OR = 1.185, 95% CI: 1.044–1.346, β2 = 0.170) and also positively associated with TELO2 (OR = 1.161, 95% CI: 1.054–1.278, β1 = 0.149). The total effect size of the exposure on outcomes (β_all = 0.124) was consistent with the mediating effect size (β12 = 0.025) and the mediating effect ratio (β12_p = 20.46%). Analysis of TELO2 in breast cancer via bioinformatics To validate TELO2 as a potential PB target in breast cancer, we analyzed the TCGA and GSE20685 datasets, finding significantly higher TELO2 expression in cancerous versus normal tissues (p < 0.05), indicating its role in cancer initiation and progression (Fig. 4A-B). ROC analysis demonstrated TELO2’s diagnostic potential with AUC values of 0.803 (TCGA) and 0.876 (GSE20685) (Fig. 4C-D). In the TCGA cohort, high TELO2 expression at nodal stages N0 and N2 was linked to poor prognosis (Fig. 4E). Clinically, TELO2 expression correlated significantly with age, gender, stage, T/N/M stage, anatomical neoplasm type, and breast cancer subtype (PAM50) (Fig. 4F-M). Further analysis of pathological sections from the HPA database supported these findings (Fig. 4N-O). Plot of bioinformatics analysis of TELO2 expression. (A) Differential expression analysis of TELO2-related genes between normal and BC in TCGA. (B) Differential expression analysis of TELO2-related genes between normal and BC in GSE20685. (C) ROC curves of TELO2 of BC in TCGA. (D) ROC curves of TELO2 of BC in GSE20685. (E) The Kaplan–Meier survival plot for TELO2 comparing N0 and N2 stages. (F-M) Box plots depict TELO2 expression differences across variables such as age, gender, tumor stage, T stage, N stage, M stage, anatomical neoplasm subtype, and PAM50 BC subtype. (N) Normal tissue of TELO2 immunohistochemical results from the HPA database. (O) Tumor tissue of TELO2 immunohistochemical results from the HPA database. * P < 0.05, ** P < 0.01, *** P < 0.001; ns, not significant TELO2 expression in breast cancer samples from the TCGA database was examined using ssGSEA immune cell infiltration analysis with high–low grouping (Fig. 5A). Distinct variations were observed in the infiltration of macrophages, mast cells, neutrophils, NK CD56bright cells, NK cells, and CD8 T cells. Correlation analysis revealed positive associations between TELO2 and NK CD56bright cells, NK cells, and CD8 T cells and negative correlations with T helper cells, macrophages, T cells, and neutrophils (Fig. 5B). CIBERSORT analysis evaluated the proportions of 22 immune cell types between high and low TELO2 groups (Fig. 5C), identifying notable differences in resting and activated memory CD4 T cells, naive B cells, resting and activated NK cells, M2 macrophages, activated dendritic cells, and resting mast cells (Fig. 5D). TELO2 was positively correlated with regulatory T cells (Tregs), resting mast cells, and activated NK cells, while showing negative correlations with M2 macrophages, activated memory CD4 T cells, and gamma delta T cells (Fig. 5E). TCGA database analysis, including single-gene differential and GSEA enrichment analyses, indicated TELO2’s involvement in pathways such as cytochrome P450-mediated drug and xenobiotic metabolism, DNA damage-induced telomere stress-related senescence, and steroid hormone biosynthesis (Fig. 5F). These findings suggest that TELO2 may influence the impact of external compounds on breast cancer pathways and functions. Immune infiltration, GSEA, and single cell analysis. (A) Box plot illustrating immune cell infiltration differences between high and low TELO2 groups based on the ssGSEA algorithm. (B) Correlation lollipop chart of TELO2 and ssGSEA immune cells. (C) Comparison of CIBERSORT immune cell proportions between high and low TELO2 groups by histogram. (E) Correlation lollipop chart of TELO2 and CIBERSORT immune cells. (F) Ridge plot of the posterior distribution of GSEA enrichment. (G) Single-cell RNA analysis of the cells using the UMAP algorithm. (H) Single-cell RNA analysis of annotated clusters by marker genes. (I) Feature plot of the expression of TELO2 concentrated in breast cancer cell subsets. * P < 0.05, ** P < 0.01, *** P < 0.001; ns, not significant Seven single-cell samples (GSM5354513, GSM5354515, GSM5354519, GSM5354520, GSM5354522, GSM5354535, GSM5354536) were randomly selected from the GSE176078 dataset in the GEO database for analysis. UMAP-based dimensionality reduction clustering identified 19 distinct cell clusters (Fig. 5G). Immune cell annotation revealed 12 primary subtypes (Fig. 5H). TELO2 expression was observed in cancer cells, endothelial cells, neutrophils, and other cell subtypes (Fig. 5I), supporting the MR mediation analysis of TELO2’s role in breast cancer via immune cells. Results of molecular docking and dynamics simulations Molecular docking analysis was performed to explore the interaction between PBs and TELO2, identified by UniProt ID Q9Y4R8, with an X-ray resolution of 2.45 Å. For each of the four compounds, five docking conformations were generated, and the conformation with the lowest binding free energy was selected for evaluation (Table 2). The binding affinity thresholds were categorized as follows: extremely weak or negligible for values above − 4 kcal/mol, moderate for values between − 7 and − 4 kcal/mol, and strong for values at or below − 7 kcal/mol. The conformation with the lowest binding free energy was chosen for detailed analysis based on the energy ranking. All docking results indicated affinities below − 4 kcal/mol, implying moderate binding strength (Fig. 6A-D). Table 2List of molecular docking List of molecular docking Molecular docking, molecular dynamics, and interaction network analysis. (A) Molecular docking result of MP and TELO2. (B) Molecular docking result of EP and TELO2. (C) Molecular docking result of PP and TELO2. (D) Molecular docking result of MP and TELO2. (E) RMSD values of small molecule ligands. (F) RMSF values of protein amino acid residues. (G) Variation in the Rg values. (H) Variation in the radius of FEL In the molecular dynamics study, the RMSD curve of the MP-TELO2 complex showed significant fluctuations in the initial kinetic phase (0–90 ns) and gradually stabilized towards equilibrium from 90 to 100 ns.6E). The root mean square fluctuation (RMSF) curve highlighted substantial conformational changes in the TELO2 protein at residues 10–11 and 111–118, suggesting increased flexibility in these regions. Additionally, distinct flexible regions were identified at specific local amino acid residue sites within the complex, which may be closely associated with molecular function and potentially influence molecular stability (Fig. 6F). The radius of gyration (Rg) curve demonstrated that after equilibration, the system exhibited minimal fluctuations, stabilizing around 2.78 nm (Fig. 6G). The free energy landscape of TELO2 (Fig. 6H) confirmed the presence of a dominant conformation and a stable structure within the Rg range of 1.42–1.98 nm and RMSD range of 2.29–2.5 nm, thus providing a foundation for investigating the characteristic conformation and stability of the MP-TELO2 complex. Molecular dynamics simulations validated the effective binding of MP to TELO2. mRNA–miRNA–IncRNA interaction network and gene interaction analysis results In the miRDB (https://mirdb.org/), DIANA-MicroT (https://dianalab.e-ce.uth.gr/microt_webserver/#/interactions), miRWalk (http://129.206.7.150/), and starBase databases (https://rnasysu.com/encori/index.php), the intersection of miRNAs appearing more than twice was analyzed, leading to the identification of 19 miRNAs associated with TELO2 and 189 lncRNAs corresponding to TELO2–miRNA interactions. The findings were used to develop an mRNA–miRNA–lncRNA interaction network (Fig. 6I). The GeneMANIA database was utilized to perform a functional analysis of the TELO2 gene and its interacting partners (Fig. 6J). The results demonstrated that TELO2 plays a role in Target of Rapamycin (TOR) signaling and protein localization to the chromosome. Results of a cross-sectional analysis The final analysis included 6,847 eligible participants (20–80 years old) from the 2005–2012 NHANES, among whom 102 breast cancer cases were identified (incidence 1.49%). Significant differences (p < 0.05) were observed between cases and controls regarding demographics, lifestyle, comorbidities, and paraben levels (MP, PP, BP), but not for EP (Table 3). Missing data exclusion induced no significant bias (p > 0.05). Limited sample size precluded significant logistic regression models (Supplementary Table 8), highlighting the need for larger studies. Table 3List of PBs analysis in breast cancerCharacteristicBreat cancerP value2Overall, N = 6847 1No, N = 6745 1Yes, N = 102 1Age48.0 (33.0, 63.0)47.0 (33.0, 63.0)68.0 (62.0, 79.0) < 0.001 Gender < 0.001 Female3459 (51%)3357 (50%)102 (100%) Male3388 (49%)3388 (50%)0 (0%)Race0.064 Mexican American1100 (16%)1089 (16%)11 (11%) Non-Hispanic Black1492 (22%)1468 (22%)24 (24%) Non-Hispanic White3114 (45%)3056 (45%)58 (57%) Other Hispanic602 (8.8%)597 (8.9%)5 (4.9%) Other Race539 (7.9%)535 (7.9%)4 (3.9%)Education0.4 Above high school3372 (49%)3330 (49%)42 (41%) Below high school794 (12%)778 (12%)16 (16%) High school2673 (39%)2629 (39%)44 (43%) Unkown8 (0.1%)8 (0.1%)0 (0%)Income0.7 Poverty1339 (20%)1321 (20%)18 (18%) Richer4935 (72%)4858 (72%)77 (75%) Unkown573 (8.4%)566 (8.4%)7 (6.9%)BMI28 (24, 32)28 (24, 32)28 (24, 34)0.3Drinker 0.002 No1723 (25%)1681 (25%)42 (41%) Unkown606 (8.9%)600 (8.9%)6 (5.9%) Yes4518 (66%)4464 (66%)54 (53%)Smoker 0.007 No4870 (71%)4790 (71%)80 (78%) Unkown361 (5.3%)351 (5.2%)10 (9.8%) Yes1616 (24%)1604 (24%)12 (12%)Hypertension < 0.001 No4479 (65%)4440 (66%)39 (38%) Unkown6 (< 0.1%)5 (< 0.1%)1 (1.0%) Yes2362 (34%)2300 (34%)62 (61%)Diabetes 0.001 Diabetes1042 (15%)1014 (15%)28 (27%) Normal3945 (58%)3902 (58%)43 (42%) Prediabetes1860 (27%)1829 (27%)31 (30%)MP71 (17, 259)71 (17, 258)105 (35, 335) 0.018 EP1 (1, 7)1 (1, 7)1 (1, 15)0.5PP9 (1, 53)9 (1, 52)21 (4, 81) < 0.001 BP0.1 (0.1, 0.8)0.1 (0.1, 0.7)0.1 (0.1, 1.7) 0.049 1Median (Q1, Q3); n (unweighted) (%)2Design-based KruskalWallis test; Pearson’s X^2: Rao & Scott adjustment List of PBs analysis in breast cancer 1Median (Q1, Q3); n (unweighted) (%) 2Design-based KruskalWallis test; Pearson’s X^2: Rao & Scott adjustment Discussion Widely used as preservatives in food, pharmaceuticals, and personal care products, PBs are being detected with increasing frequency and concentration in environmental and human samples globally. The safety of PBs is debated, with the long-term health effects from environmental exposure to them needing further study [54]. This research utilized network toxicology, bioinformatics, and Mendelian randomization to assess PBs’ potential toxicity and targets in breast cancer. Using databases such as ProTox-3.0, ADMETlab3.0, and CTD, we predicted PBs’ physicochemical properties and environmental toxicity. We identified 2851 potential PBs targets and performed GO and KEGG enrichment analyses, revealing significant biological processes, molecular functions, cellular components, and signaling pathways influenced by these compounds. Two-sample MR analysis established a causal link between 172 potential biomarkers and breast cancer risk, highlighting the association between PBs and breast cancer. TELO2 emerged as a significant risk factor for tumor development. MR-based mediation analysis showed that NKT cells significantly mediate the TELO2-breast cancer link, with a mediation proportion of 20.46%. We conducted PPI, GO, and KEGG analyses to explore the biological relevance of these targets. Bulk RNA sequencing assessed TELO2’s clinical relevance in breast cancer by analyzing disease–normal expression, prognosis, gender differences, and cancer stages, along with ssGSEA and CIBERSORT-based immune infiltration. Single-cell RNA sequencing further examined expression differences in breast cancer, endothelial cells, neutrophils, and other subtypes. GSEA revealed that PBs-treated TELO2 primarily impacts pathways related to cytochrome P450-mediated drug metabolism, xenobiotic metabolism, reactive oxygen species-induced DNA damage, telomere stress-induced senescence, and steroid hormone biosynthesis. The interaction between PBs and TELO2 suggests a mechanism linking PBs exposure, TELO2 dysregulation, and breast cancer risk. mRNA–miRNA–lncRNA network analysis provided insights into post-transcriptional regulatory mechanisms. These methods reduce confounding variables, improve causal inference reliability, and allow deeper exploration of the biological significance and mechanisms of identified targets, enhancing understanding of PBs-related health risks and their role in breast cancer pathogenesis. The study identified 172 PB targets associated with breast cancer, mainly involved in the PI3K-Akt signaling, cell cycle regulation, and PPAR signaling pathways. Dysregulation of the PI3K-Akt pathway in breast cancer promotes cell proliferation, survival, and metabolic reprogramming [55]. A key characteristic of breast cancer is the disruption of the cell cycle [56]. Activation of the PI3K-Akt pathway enhances Cyclin D1 and CDK4/6 expression, facilitating the G1 to S phase transition and boosting cell proliferation [57]. Oxidative stress from PBs may cause DNA damage and impair nucleotide excision repair, leading to genomic instability and potentially promoting breast cancer cell growth [58]. PBs might counteract PPARγ-induced cell cycle arrest by suppressing PPAR signaling [59, 60], which is crucial for regulating inflammation and cell differentiation, and its disruption is linked to breast cancer. PBs may negatively affect PPARγ, enhancing pro-inflammatory mediators and the tumor microenvironment’s oncogenic traits [61–63]. The interaction between PPARγ and the PI3K-Akt pathway suggests PBs could have dual effects [64], as inhibiting PPARγ or activating PI3K-Akt may accelerate breast cancer progression [65]. This aligns with studies showing environmental disruptors activate PI3K-Akt, boosting cancer cell survival [66]. TELO2 plays a crucial role in breast cancer by regulating the PI3K–Akt and PPARγ signaling pathways, as well as cell cycle progression. As an mTOR-interacting protein, TELO2 binds to RICTOR and enhances mTORC2 activity, thereby promoting PI3K–Akt signaling and influencing breast cancer cell proliferation, migration, and survival [67, 68]. This pathway is frequently hyperactivated in breast cancer and correlates with tumor invasiveness, poor prognosis, and regulation of cell cycle-related proteins [69]. Meanwhile, PPARγ, a nuclear receptor with tumor-suppressive potential, can inhibit the PI3K–Akt pathway and suppress tumor growth upon activation [70]. High PPARγ expression is associated with a favorable prognosis in breast cancer patients [71]. TELO2 may thus indirectly modulate PI3K–Akt signaling and cell cycle progression via PPARγ. TELO2 enrichment analysis in breast cancer samples indicates its role in epigenetic regulation. Environmental PB exposure might alter gene methylation or acetylation related to breast cancer via ERα signaling or interaction with epigenetic enzymes such as DNMTs and HDACs [72], suggesting environmental factors contribute to cancer progression through epigenetic mechanisms. Additionally, PBs may affect xenobiotic metabolism via the cytochrome P450 system in the TELO2-enriched pathway, potentially metabolizing PBs into more toxic derivatives that cause DNA damage or activate oncogenic pathways [73, 74]. Further research is needed to confirm these effects. Our study highlights a potential causal link between PBs and breast cancer, emphasizing TELO2 as a key target. MR analysis provides strong evidence that genetic variants affecting TELO2 expression may mediate the influence of PB exposure on breast cancer risk. Building on this, the established biological roles of TELO2 provide plausible mechanisms for its involvement in breast carcinogenesis. TELO2, part of the TTT complex, is crucial for breast cancer cell proliferation and survival [75]. Variations in TELO2 can destabilize the TTT complex, thereby impacting intracellular signaling and gene expression [76]. Furthermore, our pathway analysis aligns with its role in genome integrity, as TELO2 may affect telomere stability by modulating the DNA damage response, with defects potentially allowing cancer cells to evade senescence and proliferate [77]. Additionally, TELO2 is linked to the Wnt/β-catenin signaling pathway, with studies suggesting that ivermectin can inhibit this pathway by binding to TELO2, highlighting its role in breast cancer signal transduction [78]. Aberrant TELO2 expression may also disrupt DNA repair pathways, which could increase breast cancer cell sensitivity to chemotherapy [79]. Thus, TELO2 emerges from our study as a pivotal integrator of multiple oncogenic processes, potentially linking PB exposure to breast cancer development. These findings underscore the critical role of TELO2 in breast cancer initiation and progression. Mediation analysis, revealing a mediation proportion of 20.46%, further suggests that TELO2 may promote breast carcinogenesis by impairing the function of innate-like NKT cells—which regulate anti-tumor immunity through cytokine secretion such as IFN-γ and IL-4 [80, 81]. Furthermore, the interactions between NKT cells and other immune cells such as NK cells and T cells play a critical role in the immune regulation of breast cancer [82, 83]. This observation aligns with established literature indicating that NKT cell dysfunction enhances tumor progression and highlights the broader impact of the immune microenvironment on tumor behavior and patient outcomes [84]. This study has several limitations. Computational predictions from databases such as ProTox-3.0 and ADMETlab3.0 may not accurately represent in the vivo toxicity and pharmacodynamic kinetics of PBs due to reliance on existing data, potentially overlooking biological variations or interactions. This study’s NHANCE analysis is exploratory and the number of cases is limited. Therefore, the association between PBs and breast cancer still requires larger-scale prospective cohort and longitudinal cohort studies to be confirmed. Additionally, integrating multiple databases may lead to consistency issues in data interpretation, necessitating careful validation across platforms. Moreover, this study relied on computational biology and epidemiological analysis, lacking in vitro and in vivo experiments to verify the interaction between TELO2 and parahydroxybenzoates as well as the mediating role of NKT cells, in order to further confirm the causal relationship. These limitations highlight the need for further research to confirm the findings and explore the mechanisms of PBs exposure in breast cancer. Conclusions In conclusion, this integrative study demonstrates that PBs may promote breast cancer development by specifically targeting the TELO2. Our MR analysis established a robust causal link between TELO2 and breast cancer outcomes, positioning it as a key mediator of PB-induced carcinogenesis. These findings significantly advance our understanding of the toxicological mechanisms of PBs and identify TELO2 as a crucial molecular target, providing a foundation for future risk assessment and therapeutic strategies. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1: Supplementary Table 1: Potential targets of MP, EP, PP, and BP Supplementary Material 2: Supplementary Table 2: Breast cancer targets in the GeneCards database Supplementary Material 3: Supplementary Table 3: PPI topological network analysis Supplementary Material 4: Supplementary Table 4: Detailed Mendelian Randomization (MR) Analysis Results for TELO2 (Benign Breast) Supplementary Material 5: Supplementary Table 5: Detailed Mendelian Randomization (MR) Analysis Results (Breast Exallc) Supplementary Material 6: Supplementary Table 6: Mendelian randomization analysis of the causal relationship between immune cells and breast cancer IVW results Supplementary Material 7: Supplementary Table 7: Mendelian randomization analysis of gene and immune cell causality IVW results Supplementary Material 8: Supplementary Table 8: Logistic regression model analysis of PBs and breast cancer Acknowledgements The authors are grateful to Xiantao Academic and HOME for Researchers for their advanced online tools, which simplified the complex statistical analysis conducted in our research. Abbreviations Parabens Methylparaben Ethylparaben Propylparaben Butylparaben Gene Ontology Kyoto Encyclopedia of Genes and Genomes Biological process Molecular function Cellular component Protein-protein interaction Research Collaboratory for Structural Bioinformatics Protein Data Bank Phosphoinositide 3-Kinase/Protein Kinase B signaling Telomere Maintenance 2 Kaplan-Meier survival Odds ratio Hazard ratio Confidence interval Breast Cancer Mendelian Randomization Expression quantitative trait loci Instrumental variable Single Nucleotide Polymorphism Inverse-varianceweighted Area Under the Curve Receiver Operating Characteristic Overall Survival Single sample Gene Set Enrichment Analysis Cell-type Identification By Estimating Relative Subsets Of RNA Transcripts Gene Set Enrichment Analysis Root Mean Square Deviation Root Mean Square Fluctuation Radius of Gyration Free energy landscape Tumor stage Node stage Metastasis stage Uniform Manifold Approximation and Projection Author contributions Conceptualization, Xianglong Meng; Data curation, Xiaofen Li and Zhulin Bu; Formal analysis, Yuting Li and Lin Yang; Funding acquisition, Xianglong Meng; Investigation, Bin Dong; Methodology, Jing Ren, Yuting Li and Lin Yang; Project administration, Yuhui Wu and Shuosheng Zhang; Resources, Xiaofen Li; Software, Xiaofen Li; Supervision, Bin Dong and Shuosheng Zhang; Validation, Jing Ren; Visualization, Yuting Dai; Writing – original draft, Jing Ren and Huaixi Xing; Writing – review & editing, Xianglong Meng. Funding This work was supported by the Shanxi Province Science Foundation for Youths (No. 20210302124694), the Science and Technology Innovation Ability Cultivation Program Projects of Shanxi University of Chinese Medicine (Nos. 2023BK38, 2022PY-TH-02, 2022JD-KF-04, 20230034, 2023 − 156, 2023ZYYA2012, 2021L364, 202104010910029, 2022TD1014), the characterization of medicinal material properties during the “fire/temperature” manufacturing process in Shanxi Province (CZ2023041_019), a TCM resource survey (CZ2023016), the Shanxi Key Laboratory of Chinese Medicinal Materials Processing (202104010910029), the Shanxi Modern Agricultural Industry Traditional Chinese Medicine Technology System (CZ2023018), the Jinyao Comprehensive Development and Utilization Collaborative Innovation Center (CZ2023056_003, XN2024040_006), pharmaceutical processing discipline construction at Shanxi University of Chinese Medicine (2023XKJS-22), the Shanxi Key Laboratory of Traditional Chinese Medicine Processing (XN2024040_006), TCM processing discipline construction (XN2024093_024), the Chinese Medicine - Key Research Laboratory (CZ2024010_002), and the Taiyuan City unveiling-hanging project on quality improvement and industrialization of Shanxi authentic Chinese medicine slices (KY2024102). Data availability The original contributions presented in this study are included in the article/Supplementary Materials. Further inquiries can be directed to the corresponding author. Declarations Ethics approval and consent to participate Not applicable. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1: Supplementary Table 1: Potential targets of MP, EP, PP, and BP Supplementary Material 2: Supplementary Table 2: Breast cancer targets in the GeneCards database Supplementary Material 3: Supplementary Table 3: PPI topological network analysis Supplementary Material 4: Supplementary Table 4: Detailed Mendelian Randomization (MR) Analysis Results for TELO2 (Benign Breast) Supplementary Material 5: Supplementary Table 5: Detailed Mendelian Randomization (MR) Analysis Results (Breast Exallc) Supplementary Material 6: Supplementary Table 6: Mendelian randomization analysis of the causal relationship between immune cells and breast cancer IVW results Supplementary Material 7: Supplementary Table 7: Mendelian randomization analysis of gene and immune cell causality IVW results Supplementary Material 8: Supplementary Table 8: Logistic regression model analysis of PBs and breast cancer Data Availability Statement The original contributions presented in this study are included in the article/Supplementary Materials. Further inquiries can be directed to the corresponding author.