Integrative pan-cancer analysis of transferrin reveals context-dependent prognostic associations and links to immune and metabolic disease-related programs Authors: Weiyi Kong, Ranran Zhang, Zhenyang Fu, Sicheng Chen, Yining Dai, Haomin Xie, Zongrong Lin, Hongcheng Chen, He Li, Yu Chen Published: 2026 ABSTRACT BackgroundIron metabolism is closely linked to tumor biology, yet the pan-cancer significance of transferrin (TF), the major circulating iron-transport protein, remains insufficiently defined. Although TF has been implicated in cancer-related processes, its prognostic relevance, immune associations, and broader disease-related transcriptional context have not been systematically characterized across tumor types.ObjectiveThis study aimed to perform an integrative pan-cancer analysis of TF to characterize its expression patterns, clinical associations, immune context, pathway features, and pharmacogenomic correlations, and to explore whether TF-related signals extend to selected metabolic and chronic organ injury settings.MethodsWe used multiple public databases, including The Cancer Genome Atlas (TCGA), Human Protein Atlas (HPA), Gene Expression Omnibus (GEO), and Cancer Cell Line Encyclopedia (CCLE), to integrate transcriptomic, proteomic, and clinical data across 33 tumor types and selected non-malignant conditions. TF expression was evaluated across normal tissues, tumors, and cell lines, followed by survival analysis, immune infiltration analysis, TMB/MSI and methylation assessment, pathway enrichment, and drug-response correlation. Independent GEO cohorts of non-alcoholic steatohepatitis (NASH), heart failure (HF), and liver cirrhosis (LC) were used for cross-disease extension. Selected findings were further explored in OA/PA-treated hepatocytes, 786-O renal carcinoma cells, and AC16 cardiomyocytes.ResultsTF showed pronounced tissue specificity and cancer-type-dependent dysregulation. Across pan-cancer cohorts, the most consistent adverse survival associations were observed in kidney renal clear cell carcinoma (KIRC) and stomach adenocarcinoma (STAD), where TF remained associated with overall survival (OS) in multivariable analyses. TF expression was also correlated with cancer-type-specific immune infiltration patterns and selected drug-response profiles. Across independent NASH, HF, and LC datasets, TF expression was elevated and TF-associated pathways partially overlapped with those observed in cancer. In vitro experiments provided preliminary support that TF modulation is associated with proliferative phenotypes in KIRC cells and stress- and metabolism-related phenotypes in hepatocyte and cardiomyocyte models.ConclusionThese findings support TF as a context-dependent biomarker candidate in cancer, with the most consistent prognostic relevance observed in KIRC and STAD. Rather than establishing a unified mechanism across diseases, this study provides an integrative framework suggesting that TF is associated with malignant behavior, immune context, and selected metabolic stress-related programs, and warrants further mechanistic investigation.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1007/s12672-026-05313-5. Introduction Iron metabolism is essential for maintaining cellular balance, as it contributes to DNA synthesis, oxygen transport, and energy production [1]. TF, a major iron-binding glycoprotein in plasma, plays a key role in systemic iron delivery and recycling [2]. Recent research has revealed that tumor progression, cancer cell proliferation, oxidative stress responses, and immune evasion are closely related to the dysfunction of iron balance [3]. Even with growing interest in this field, existing pan-cancer studies have focused on intracellular iron storage or membrane-bound receptors. A systematic examination of TF as a systemic iron carrier across different cancers and related diseases remains lacking. Recent evidence suggests that TF may also be relevant to the tumor microenvironment (TME), where altered iron handling has been linked to immune context, tumor progression, and therapy response [4]. TF is functionally linked to transferrin receptor-mediated iron uptake and related iron-regulatory pathways. Its detection in exosomes and blood microparticles indicates possible functions beyond iron transport, including the modulation of angiogenesis, inflammation, and coagulation-processes relevant to cancer development and metastasis [5]. Given the established role of TF in systemic iron transport and the growing interest in iron-dependent processes in cancer, we performed an integrative pan-cancer analysis to characterize TF across multiple tumor types using publicly available transcriptomic, proteomic, and clinical datasets [6]. Rather than assuming a uniform oncogenic role, we asked whether TF shows cancer-type-specific patterns of expression, prognostic association, immune context, and regulatory features, and whether these patterns extend to selected non-malignant disease settings. We therefore combined pan-cancer analyses with exploratory validation in NASH, HF, and LC datasets [7], and supplemented these findings with targeted in vitro experiments in KIRC, hepatocyte, and cardiomyocyte models. Through this approach, we aimed to provide a more precise and restrained assessment of TF as a candidate biomarker and biologically relevant factor in cancer and related metabolic stress contexts. Materials and methods Public datasets and preprocessing Publicly available datasets were used for the integrative analysis of TF across cancers and selected non-malignant disease contexts. Pan-cancer transcriptomic and clinical data were obtained from TCGA, with additional normal tissue reference data from the Genotype-Tissue Expression (GTEx) project. Protein expression and representative immunohistochemistry images were retrieved from the HPA. Cancer cell line expression data were obtained from the CCLE [8], the Genomics of Drug Sensitivity in Cancer (GDSC) [9, 10], and the NCI-60 panel. External disease validation datasets for NASH, HF, and LC were downloaded from the GEO [11]. Relevant pharmacogenomic data were obtained from GDSC and the Cancer Therapeutics Response Portal (CTRP) [12]. Public resources used for selected supplementary analyses are summarized in Supplementary Materials 2: Dataset links. For TCGA-based analyses, harmonized RNA-seq data [13] were processed using R (v4.0.3) [14, 15, 16] and normalized to transcripts per million (TPM). For survival analyses, the uniformly processed TCGA TARGET GTEx (PANCAN) cohort from UCSC Xena was used. Samples with follow-up < 30 days were excluded. Cancer types with insufficient sample size for a given survival endpoint were also excluded. The final numbers of evaluable cancer types for OS, DSS, PFI, and DFI were as reported in the Results. Additional preprocessing details, platform-specific processing information, and software/version tables are provided in Supplementary Materials 2: Table S1. Expression and survival analyses TF expression was assessed across normal tissues, tumor tissues, and cell lines using TCGA, GTEx, HPA, CCLE, GDSC, and NCI-60 resources [17]. We performed tumor vs. normal tissue and paired tumor vs. adjacent normal tissue comparisons for all datasets with corresponding available data [18]. We validated protein-level expression patterns using tissue expression data and representative immunohistochemistry (IHC) images from the HPA. The clinical information of 33 tumors was downloaded from the TCGA database. We selected OS, DSS, DFI and PFI to analyze the correlation between TF expression and the 33 cancers’ prognoses. For Kaplan-Meier analyses, the patients were separated into high and low TF groups according to the median TF expression value with 50% cut-off high and 50% cut-off low. Log-rank tests were used to compare survival distributions [19, 20]. To validate the independent prognostic value of TF, we further performed multivariable Cox proportional hazards regression analysis for specific tumor types with consistent and significant prognostic associations in univariate survival analyses, by adjusting for common available clinical covariates including age, gender, tumor stage, and histological grade. To further explore the predictive power of TF, a time-dependent Receiver Operating Characteristic (ROC) curve was also conducted. As supplementary exploratory and validation analyses, we additionally performed: (1) diagnostic ROC analyses to evaluate the tumor diagnostic value of TF; (2) correlation analyses to investigate the association between TF expression and tumor pathological stage; (3) head-to-head comparisons of the prognostic and diagnostic performance between TF and canonical iron metabolism-related reference genes. All related results are presented in Supplementary Material 1, Figures S1–S2 and S5. Immune infiltration, TMB/MSI, and methylation analyses To investigate the regulatory association between TF expression and the TME, we assessed the correlation between TF expression and immune cell infiltration levels across pan-cancer cohorts from TCGA, leveraging multiple well-established transcriptome-based immune deconvolution algorithms integrated within the TIMER2.0 platform [21]. To investigate the regulatory role of TF in the TME, we performed comprehensive immune infiltration profiling via single-sample gene set enrichment analysis (ssGSEA). We subsequently performed correlation analyses to evaluate the correlation between TF expression and ssGSEA-derived immune cell infiltration levels. Tumor mutational burden (TMB) is defined as the total number of base mutations per million cells in the tumor [22]. Microsatellite instability (MSI) is caused by abnormal DNA MMR and leads to gene duplication disorder and tumor development, affecting tumor prognosis [23]. Correlation between TF and TMB and MSI was conducted using Spearman correlation analysis. The major types of eukaryotic RNA methylation include N6-methyladenosine (m6A), 5-methylcytidine (m5C), and N1-methyladenosine (m1A). We evaluated the correlation between TF expression and key regulatory proteins of m6A, m5C, and m1A modifications via Spearman’s rank correlation analysis. Analyses of molecular subtypes, immune subtypes, and promoter methylation [24] were treated as supplementary contextual analyses and are reported in Supplementary Materials 1: Figures S3–S4. Functional enrichment and drug sensitivity analyses To characterize TF-associated biological programs, we performed co-expression and network-based analyses using publicly available transcriptomic datasets and curated protein interaction resources. Protein-protein interaction (PPI) networks were constructed using STRING. The top TF-associated genes were used for Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses [25, 26]. GSEA was performed to identify pathway-level differences associated with TF expression states. We performed drug sensitivity analyses using the GDSC and CTRP datasets to assess the correlation between TF expression and drug sensitivity profiles across cancer cell lines. Candidate compounds with significant associations were summarized in the main figures, whereas exploratory analyses related to predicted TF regulators, including candidate miRNAs [27–29, 30], transcription factor networks, and RNA modification-related regulators [31], were presented in Supplementary Materials 1: Figure S6. External validation in NASH, HF, and LC datasets To extend the analysis beyond cancer, TF expression was examined in independent GEO datasets representing NASH, HF, and LC. We assessed differential expression between disease and control tissues in each included dataset and performed ROC analyses to evaluate the diagnostic discriminatory power of TF expression [32, 33]. In addition, we conducted integrated pathway enrichment analyses of TF-associated gene signatures in these disease cohorts using GSEA and enrichment approaches to evaluate whether TF-associated transcriptional programs in cancer partially overlapped with those identified in specific metabolic or chronic organ injury-related conditions [34, 35]. The detailed pathway plots and expanded enrichment results are provided in Supplementary Materials 1: Figure S7. Cell models and in vitro assays The human normal hepatocyte cell line L02 and the human renal cortical proximal tubule epithelial cell line HK-2 were obtained from the Cell Bank of the Chinese Academy of Sciences. The human renal clear cell carcinoma cell line 786-O and the human cardiomyocyte cell line AC16 were obtained from the American Type Culture Collection. Cells were cultured in the media recommended for each line and maintained at 37℃ in a humidified atmosphere containing 5% CO2. Regular mycoplasma testing was performed. To generate a lipotoxic hepatocyte model, L02 cells were treated with an oleic acid/palmitic acid (OA/PA) mixture for 24 h, with bovine serum albumin vehicle-treated cells as controls. We generated stable TF overexpression in target cells through a recombinant lentiviral expression system, and achieved transient TF knockdown via transfection of specific small interfering RNA (siRNA) targeting TF. Expression changes were validated by qPCR and Western blotting. Functional assays included cell viability, EdU incorporation, colony formation, migration, invasion, lipid accumulation, triglyceride measurement, reactive oxygen species detection, mitochondrial membrane potential assessment, and apoptosis-related analyses, according to the experimental requirements of each model. Detailed information regarding reagents, antibodies, oligonucleotide sequences, assay-specific experimental protocols, and unprocessed original western blot images are available in the Supplementary Material, with the full raw blot data presented in Supplementary Figure S8. Statistical analysis Bioinformatic analyses were performed primarily in R, and experimental data were analyzed using GraphPad Prism (version 10.2). For continuous variables, we used an unpaired two-tailed Student’s t-test for comparisons between two independent groups, and one-way analysis of variance (ANOVA) followed by Tukey’s post hoc test for pairwise comparisons across three or more groups. For correlation analyses, Pearson’s correlation coefficient was applied for normally distributed continuous variables, while Spearman’s rank correlation coefficient was used for variables with non-normal distribution. For high-dimensional omics analyses, we applied the Benjamini-Hochberg (BH) method for multiple-testing correction, with a false discovery rate (FDR) < 0.05 set as the threshold for statistical significance. For conventional single-variable hypothesis tests with low-dimensional data, statistical significance was defined as a two-tailed P value < 0.05. Quantitative experimental data are presented as mean±standard deviation (SD) from at least three independent experiments unless otherwise indicated. Detailed software/version information, figure-to-platform mapping, and supplementary analysis workflows are provided in Supplementary Materials 2: Table S1. Results Fig. 1Study design and analytical framework for the integrative characterization of TF across cancers and selected metabolic disease contexts. The overall study design is summarized in Fig. 1. We first characterized the expression profile of TF across normal tissues, tumors, and cancer cell lines using public multi-omics resources. We then evaluated its prognostic associations across pan-cancer cohorts, followed by analyses of TF-associated functional pathways, immune context, and drug-response patterns. To extend the analysis beyond cancer, we examined TF expression and TF-associated transcriptional features in independent datasets of non-alcoholic steatohepatitis (NASH), heart failure (HF), and liver cirrhosis (LC). Finally, selected bioinformatic findings were explored in KIRC, hepatocyte, and cardiomyocyte cell models Study design and analytical framework for the integrative characterization of TF across cancers and selected metabolic disease contexts. The overall study design is summarized in Fig. 1. We first characterized the expression profile of TF across normal tissues, tumors, and cancer cell lines using public multi-omics resources. We then evaluated its prognostic associations across pan-cancer cohorts, followed by analyses of TF-associated functional pathways, immune context, and drug-response patterns. To extend the analysis beyond cancer, we examined TF expression and TF-associated transcriptional features in independent datasets of non-alcoholic steatohepatitis (NASH), heart failure (HF), and liver cirrhosis (LC). Finally, selected bioinformatic findings were explored in KIRC, hepatocyte, and cardiomyocyte cell models Fig. 2Tissue-specific expression and cancer-associated dysregulation of TF across normal tissues, tumors, and cell lines. (A) RNA expression (nTPM) and protein expression scores of TF across human tissues from HPA. (B) Comparison of TF protein expression across different normal tissues. (C) TF mRNA expression across normal and tumor-related tissue contexts. (D) TF expression across cancer cell lines from public cell line resources, ranked by RNA expression level. (E) Differential TF expression across TCGA cancer types visualized using TIMER2. (F) Paired comparison of TF expression between tumor and matched adjacent normal tissues. (G) TF expression across the NCI-60 tumor cell line panel. (H) TF expression across representative normal-derived cell lines. (I) Representative subcellular localization pattern of TF from HPA. (J-K) Representative immunohistochemistry images showing TF protein expression in normal liver tissue (J) and normal colon tissue (K) Tissue-specific expression and cancer-associated dysregulation of TF across normal tissues, tumors, and cell lines. (A) RNA expression (nTPM) and protein expression scores of TF across human tissues from HPA. (B) Comparison of TF protein expression across different normal tissues. (C) TF mRNA expression across normal and tumor-related tissue contexts. (D) TF expression across cancer cell lines from public cell line resources, ranked by RNA expression level. (E) Differential TF expression across TCGA cancer types visualized using TIMER2. (F) Paired comparison of TF expression between tumor and matched adjacent normal tissues. (G) TF expression across the NCI-60 tumor cell line panel. (H) TF expression across representative normal-derived cell lines. (I) Representative subcellular localization pattern of TF from HPA. (J-K) Representative immunohistochemistry images showing TF protein expression in normal liver tissue (J) and normal colon tissue (K) We first mapped the expression distribution of TF across normal human tissues and pan-cancer contexts to characterize the physiological expression pattern of TF and provide a baseline for subsequent tumor differential expression analysis. As shown in Fig. 2A and B, TF displayed significant tissue specificity at both the transcript and protein levels, with relatively high expression in normal liver tissues and low expression in the majority of other normal organs. This tissue-restricted expression pattern was further validated by representative IHC staining images from HPA, which showed moderate TF positivity in normal liver tissue and markedly lower staining intensity in normal colon tissue (Fig. 2J and K). Building on our findings of TF’s tissue-specific expression in normal human tissues, we next characterized its expression landscape across TCGA pan-cancer cohorts. As shown in Fig. 2C and F, TF expression varied substantially across different cancer types. TCGA-based comparisons showed differential TF expression between tumor and normal tissues in multiple cancers, although the direction and magnitude of change were not uniform. Matched tumor-adjacent paired analyses further indicated that TF dysregulation was cancer-type dependent rather than a universal, unidirectional change across all pan-cancer types (Fig. 2F). These findings suggest that aberrant TF expression occurs in several malignancies but remains strongly influenced by tissue context. Cancer cell line datasets showed a similar pattern. TF expression was highest in hepatoma-derived cell lines, whereas many non-hepatic cell lines showed relatively low to near-undetectable expression (Fig. 2D, G and H). In addition, subcellular localization data from HPA revealed detectable TF signal in the nucleoplasm and midbody in specific cellular contexts (Fig. 2I), although the functional significance of this observation falls outside the scope of the current study. Taken together, these data show that TF is not uniformly elevated across cancers, but instead displays pronounced tissue specificity together with cancer-type-dependent dysregulation. Fig. 3Context-specific prognostic associations of TF across pan-cancer cohorts, with prominent signals in KIRC and STAD. (A) Forest plot of univariate Cox regression analysis of TF for overall survival (OS) across TCGA cancer types. (B) Forest plot of univariate Cox regression analysis of TF for disease-free interval (DFI). (C) Forest plot of univariate Cox regression analysis of TF for progression-free interval (PFI). (D) Forest plot of univariate Cox regression analysis of TF for disease-specific survival (DSS). Red points indicate statistically significant associations. (E-H) Kaplan-Meier curves for overall survival in representative cancer types stratified by TF expression. (I-L) Kaplan-Meier curves for disease-free survival in representative cancer types based on GEPIA2. (M) Multivariable Cox regression analysis of overall survival in KIRC. (N) Multivariable Cox regression analysis of overall survival in STAD Context-specific prognostic associations of TF across pan-cancer cohorts, with prominent signals in KIRC and STAD. (A) Forest plot of univariate Cox regression analysis of TF for overall survival (OS) across TCGA cancer types. (B) Forest plot of univariate Cox regression analysis of TF for disease-free interval (DFI). (C) Forest plot of univariate Cox regression analysis of TF for progression-free interval (PFI). (D) Forest plot of univariate Cox regression analysis of TF for disease-specific survival (DSS). Red points indicate statistically significant associations. (E-H) Kaplan-Meier curves for overall survival in representative cancer types stratified by TF expression. (I-L) Kaplan-Meier curves for disease-free survival in representative cancer types based on GEPIA2. (M) Multivariable Cox regression analysis of overall survival in KIRC. (N) Multivariable Cox regression analysis of overall survival in STAD We selected OS, DSS, DFI, and PFI to investigate the clinical relevance of TF in pan-cancer, which revealed that the prognostic value of TF was highly heterogeneous across cancer types. High TF expression was significantly associated with adverse survival outcomes only in a limited subset of malignancies, rather than exerting a universal unfavorable prognostic effect across all cancer types (Fig. 3A and D). Among the cancers analyzed, KIRC and STAD showed the most consistent adverse prognostic associations. Kaplan-Meier analyses demonstrated shorter survival in patients with higher TF expression in both cancer types (Fig. 3E and L). In multivariable Cox regression models, TF remained significantly associated with OS in KIRC and STAD after adjustment for available clinical variables (Fig. 3M and N), supporting its potential prognostic relevance in these two malignancies. To further place these findings in context, we examined several related clinical and molecular features. First, diagnostic ROC analyses showed that the diagnostic discriminatory performance of TF varied across tumor types, and was only modest in KIRC and STAD — the two cancers where its prognostic associations were the most consistent (Supplementary Materials 1: Figure S1). Additionally, we found no consistent correlation between TF expression and tumor pathological stage in either KIRC or STAD (Supplementary Materials 1: Figure S2). These findings suggest that the prognostic associations of TF are unlikely to be driven by its tumor diagnostic capacity, nor can they be solely attributed to differences in tumor burden. We also analyzed TF expression across molecular and immune subtype classifications. TF levels differed across subtype groups in several cancers, including KIRC and STAD (Supplementary Materials 1: Figure S3), indicating that TF-associated biology may vary both between tumor types and within individual cancers. We further compared TF with representative iron metabolism-related genes. In KIRC and STAD, TF showed relatively consistent adverse survival associations compared with several related markers (Supplementary Materials 1: Figure S5), supporting its prioritization for further analysis in these cancers. Overall, these data support a context-dependent prognostic role for TF, with the strongest evidence observed in KIRC and STAD. Fig. 4Functional interaction networks and pathway enrichment analyses of TF-associated genes. (A) STRING physical interaction subnetwork centered on TF, showing representative interacting proteins involved in iron handling and related biological processes. (B) GSEA of pathways associated with TF expression. Representative pathways with positive or negative enrichment trends are shown. (C) KEGG pathway enrichment analysis of TF and its co-expressed genes, highlighting pathways related to complement and coagulation cascades, cholesterol metabolism, and PPAR signaling. (D) GO enrichment analysis of TF-associated genes, including biological process (BP), cellular component (CC), and molecular function (MF) terms. (E) Expanded STRING network showing functional interactions among TF-associated proteins. (F) Protein-protein interaction network of representative TF co-expressed genes identified from STRING-based analysis Functional interaction networks and pathway enrichment analyses of TF-associated genes. (A) STRING physical interaction subnetwork centered on TF, showing representative interacting proteins involved in iron handling and related biological processes. (B) GSEA of pathways associated with TF expression. Representative pathways with positive or negative enrichment trends are shown. (C) KEGG pathway enrichment analysis of TF and its co-expressed genes, highlighting pathways related to complement and coagulation cascades, cholesterol metabolism, and PPAR signaling. (D) GO enrichment analysis of TF-associated genes, including biological process (BP), cellular component (CC), and molecular function (MF) terms. (E) Expanded STRING network showing functional interactions among TF-associated proteins. (F) Protein-protein interaction network of representative TF co-expressed genes identified from STRING-based analysis To explore the biological context associated with TF expression, we performed interaction network and pathway enrichment analyses. In STRING-based physical interaction analysis, TF was connected to canonical iron-related proteins including TFRC, TFR2, HFE, and HJV (Fig. 4A), consistent with its established role in systemic iron transport and iron handling. The expanded STRING network further linked TF to proteins involved in extracellular transport, circulating protein systems, and related metabolic processes (Fig. 4E and F). GSEA identified one significantly negatively enriched pathway, taurine and hypotaurine metabolism, whereas several other pathways, including spliceosome, proteasome, nucleotide excision repair, and basal transcription factors, showed positive enrichment trends (Fig. 4B). These findings suggest that TF-associated expression profiles may coincide with multiple cellular programs, although with variable enrichment strength across gene sets. KEGG analysis of TF and its co-expressed genes highlighted pathways related to complement and coagulation cascades, cholesterol metabolism, and PPAR signaling (Fig. 4C). GO analysis similarly identified enrichment in biological processes and cellular components related to hemostasis, blood coagulation, blood microparticles, and extracellular matrix-associated compartments, together with molecular functions such as enzyme inhibitor activity and endopeptidase regulator activity (Fig. 4D). Rather than indicating a single TF-specific mechanism, these findings place TF within broader transcriptional programs linked to metabolism and extracellular remodeling. Taken together, the network and enrichment analyses support a biologically plausible TF-associated context involving iron-related functions together with extracellular, metabolic, and circulating protein-associated pathways. Fig. 5Associations of TF expression with tumor mutational burden, microsatellite instability, and immune infiltration across cancers. (A) Correlation heatmap between TF expression and tumor mutational burden (TMB) across TCGA cancer types. (B) Correlation heatmap between TF expression and microsatellite instability (MSI) across TCGA cancer types. (C) Pan-cancer associations between TF expression and immune cell infiltration estimated using multiple deconvolution algorithms in TIMER2. (D) Correlation matrix of TF expression and immune infiltration scores across tumor types and immune cell populations. (E) Correlations between TF expression and immune cell infiltration in KIRC. (F) Correlations between TF expression and immune cell infiltration in STAD Associations of TF expression with tumor mutational burden, microsatellite instability, and immune infiltration across cancers. (A) Correlation heatmap between TF expression and tumor mutational burden (TMB) across TCGA cancer types. (B) Correlation heatmap between TF expression and microsatellite instability (MSI) across TCGA cancer types. (C) Pan-cancer associations between TF expression and immune cell infiltration estimated using multiple deconvolution algorithms in TIMER2. (D) Correlation matrix of TF expression and immune infiltration scores across tumor types and immune cell populations. (E) Correlations between TF expression and immune cell infiltration in KIRC. (F) Correlations between TF expression and immune cell infiltration in STAD We next examined whether TF expression was associated with genomic instability-related features and tumor immune context across cancers. Pan-cancer correlation analyses showed that the relationship between TF and TMB or MSI was generally weak and heterogeneous across tumor types (Fig. 5A and B). In KIRC, TF expression showed no clear association with TMB or MSI, whereas weak but statistically significant correlations were observed in STAD. These findings suggest that the prognostic relevance of TF in selected cancers is unlikely to be explained primarily by mutational burden or microsatellite instability. Immune infiltration analyses further highlighted substantial cancer-type specificity. Across tumors, TF expression showed variable correlations with different immune cell populations depending on the cancer type and deconvolution method used (Fig. 5C and D). Positive associations were observed in some settings, whereas negative associations were seen in others, indicating that TF-related immune patterns are not uniform across cancers. In KIRC, TF expression was positively associated with plasmacytoid dendritic cells and effector memory T cells, whereas it showed no statistically significant correlations with CD8+ T cells (Fig. 5E). In STAD, TF expression showed negative correlations with several immune populations, including CD8+ T cells, macrophages, Th1 cells, Th17 cells, and NK CD56dim cells (Fig. 5F). These contrasting patterns suggest that TF expression may be linked to distinct tumor microenvironment contexts in KIRC and STAD rather than a common immune phenotype across tumor types. Promoter methylation analyses provided additional context for TF regulation. In KIRC, increased promoter methylation coexisted with relatively high TF expression, whereas no clear methylation difference was observed in STAD (Supplementary Materials 1: Figure S4). These results suggest that TF expression is likely regulated by multiple mechanisms, and that promoter methylation alone does not account for the observed expression patterns in all cancers. Overall, these analyses support an association between TF expression and tumor immune context, but do not establish direct immunoregulatory or causal effects. Fig. 6Associations between TF expression and selected drug-response profiles in pharmacogenomic datasets. (A) Correlation analysis between TF expression and drug-response metrics in the CTRP dataset. (B) Representative clinically relevant compounds associated with TF expression in CTRP. (C) Correlation analysis between TF expression and drug-response metrics in the GDSC dataset. (D) Representative clinically relevant compounds associated with TF expression in GDSC Associations between TF expression and selected drug-response profiles in pharmacogenomic datasets. (A) Correlation analysis between TF expression and drug-response metrics in the CTRP dataset. (B) Representative clinically relevant compounds associated with TF expression in CTRP. (C) Correlation analysis between TF expression and drug-response metrics in the GDSC dataset. (D) Representative clinically relevant compounds associated with TF expression in GDSC To further explore the clinical translational potential of TF, we used CTRP and GDSC to investigate the association between TF expression and tumor drug sensitivity. The results showed that TF expression had significant correlations with multiple compounds, although the specific patterns varied between the two datasets (Fig. 6A and D). Several associated compounds belonged to drug classes targeting the proteasome, MAPK-related signaling, and epigenetic regulation. These findings suggest that TF expression is associated with distinct pharmacologic response profiles across datasets, although these findings alone do not establish TF-dependent therapeutic vulnerability. As these pharmacogenomic analyses were based on cross-cell-line correlations, we interpreted them as prioritization signals rather than direct evidence of mechanism. Nevertheless, the recurrence of certain drug classes across datasets indicates that TF-associated transcriptional states may be accompanied by recognizable treatment-response patterns. To further explore possible upstream regulation of TF, we summarized predicted miRNA regulators, transcription factor interactions, and correlations with RNA modification-related regulators in supplementary analyses (Supplementary Materials 1: Figure S6). These results provide additional context for TF-associated regulatory features, while remaining computational and exploratory in nature. Fig. 7External validation of altered TF expression and diagnostic performance in NASH, HF, and liver cirrhosis cohorts. (A) TF mRNA expression in NASH and control liver tissues in GSE164760. (B) TF mRNA expression in HF and control cardiac tissues in GSE135055. (C) TF mRNA expression in LC and control liver tissues in GSE139602. (D-F) ROC curves evaluating the discriminatory performance of TF expression in GSE164760 (D), GSE135055 (E), and GSE139602 (F). (G) TF mRNA expression in NASH and control liver tissues in GSE89632. (H) TF mRNA expression in HF and control cardiac tissues in GSE21610. (I) TF mRNA expression in LC and control liver tissues in GSE14323. (J-L) ROC curves evaluating the discriminatory performance of TF expression in GSE89632 (J), GSE21610 (K), and GSE14323 (L) External validation of altered TF expression and diagnostic performance in NASH, HF, and liver cirrhosis cohorts. (A) TF mRNA expression in NASH and control liver tissues in GSE164760. (B) TF mRNA expression in HF and control cardiac tissues in GSE135055. (C) TF mRNA expression in LC and control liver tissues in GSE139602. (D-F) ROC curves evaluating the discriminatory performance of TF expression in GSE164760 (D), GSE135055 (E), and GSE139602 (F). (G) TF mRNA expression in NASH and control liver tissues in GSE89632. (H) TF mRNA expression in HF and control cardiac tissues in GSE21610. (I) TF mRNA expression in LC and control liver tissues in GSE14323. (J-L) ROC curves evaluating the discriminatory performance of TF expression in GSE89632 (J), GSE21610 (K), and GSE14323 (L) To determine whether TF-associated signals observed in cancer might also be present in non-malignant disease settings, we analyzed six GEO datasets representing NASH, HF, and LC. Across these cohorts, TF transcript levels were consistently higher in disease samples than in corresponding controls, although the magnitude of change varied between datasets (Fig. 7A, C, G and I). ROC analyses showed that TF had moderate discriminatory performance in these datasets, with AUC values demonstrating measurable but not uniform separation between disease and control samples (Fig. 7D, F, J and L). These findings extend our previous observations in malignancies, confirming that dysregulated TF expression is not restricted to tumor tissues, and may also accompany the pathological process of selected chronic metabolic or organ injury-related conditions. However, the diagnostic implications of these observations should be interpreted cautiously because of differences in cohort composition, tissue source, and study design across public datasets. To extend these observations beyond differential expression, we performed pathway-level analyses in the same disease cohorts. TF-associated genes in NASH, HF, and LC were recurrently enriched in pathways related to extracellular matrix organization, metabolism, mitochondrial function, and stress-responsive programs (Supplementary Materials 1: Figure S7). Some of these categories partially overlapped with pathways identified in cancer analyses, suggesting convergence at the level of transcriptional context rather than a shared TF-driven mechanism across diseases. Taken together, these analyses extend TF-associated observations beyond cancer datasets and suggest that TF-related expression programs may also accompany selected metabolic and chronic organ injury states. Fig. 8Experimental validation of TF-associated phenotypes in KIRC, hepatocyte, and cardiomyocyte cell models. (A) Validation of TF expression in OA/PA-treated L02 hepatocytes and 786-O KIRC cells by qPCR and Western blotting. (B) Effects of TF overexpression or knockdown on cell viability, DNA synthesis, colony formation, migration, and invasion in 786-O cells. (C) Effects of TF modulation on lipid accumulation, triglyceride content, and lipid metabolism-related gene expression in OA/PA-treated L02 cells. (D) Effects of TF modulation on reactive oxygen species, mitochondrial membrane potential, apoptosis-related readouts, and cleaved Caspase-3 in AC16 cardiomyocyte cells Experimental validation of TF-associated phenotypes in KIRC, hepatocyte, and cardiomyocyte cell models. (A) Validation of TF expression in OA/PA-treated L02 hepatocytes and 786-O KIRC cells by qPCR and Western blotting. (B) Effects of TF overexpression or knockdown on cell viability, DNA synthesis, colony formation, migration, and invasion in 786-O cells. (C) Effects of TF modulation on lipid accumulation, triglyceride content, and lipid metabolism-related gene expression in OA/PA-treated L02 cells. (D) Effects of TF modulation on reactive oxygen species, mitochondrial membrane potential, apoptosis-related readouts, and cleaved Caspase-3 in AC16 cardiomyocyte cells We next conducted targeted in vitro experiments to determine the functional consequences of TF modulation in select cellular contexts. Consistent with our bioinformatic findings, both OA/PA-treated L02 hepatocytes and 786-O KIRC cells exhibited elevated TF mRNA and protein levels relative to their respective controls (Fig. 8A). Uncropped western blot images are presented in Supplementary Materials 1: Figure S8. In 786-O cells, enforced TF expression was associated with increased cell viability, DNA synthesis, colony formation, migration, and invasion. Conversely, TF depletion suppressed these malignant traits (Fig. 8B). These findings support a functional association between TF expression and aggressive KIRC cellular phenotypes in vitro, but they do not establish a definitive causal pathway. In hepatocytes challenged with OA/PA, TF overexpression was associated with increased lipid droplet accumulation and triglyceride content, together with expression changes in genes related to lipid synthesis and fatty acid oxidation. In contrast, TF knockdown showed the opposite direction of change (Fig. 8C). These results suggest that TF modulation was associated with altered metabolic phenotypes in vitro. In AC16 cardiomyocytes, TF upregulation was related to increased reactive oxygen species levels, reduced mitochondrial membrane potential, and higher apoptosis-related readouts, including elevated cleaved Caspase-3. Notably, TF knockdown conferred partial protection against these stress-induced alterations (Fig. 8D). These findings suggest that TF modulation may influence stress-related phenotypes in cardiomyocyte cells under the conditions tested, but they do not establish direct in vivo effects on cardiac dysfunction. Overall, these experimental findings align with our computational analyses and suggest that TF is associated with proliferative and stress-related phenotypes across distinct cancer and non-cancer cell models. Discussion In this study, we performed an integrative pan-cancer analysis of TF across over 30 tumor types and extended the analysis to selected metabolic and chronic organ injury contexts. Overall, the findings support three main observations. First, TF expression exhibits pronounced tissue dependency and is not uniformly perturbed across cancers. Second, its prognostic relevance is context-specific, with the most consistent adverse survival associations observed in KIRC and STAD. Third, TF expression correlates with distinct immunological, pharmacogenomic, and pathway‑level contexts, and some of these transcriptional signatures extend to non‑malignant disease datasets. Taken together, our findings support TF as a context-dependent biomarker candidate rather than a universal cancer marker. The expression analyses should be interpreted in the context of TF biology. As the primary circulating iron carrier and a protein highly enriched in hepatobiliary tissues under physiological conditions [36], TF is expected to show strong tissue specificity [37]. This baseline distribution likely contributes to its relatively high expression in liver-associated tissues and hepatoma cell lines. However, tissue-enriched expression alone does not establish clinical relevance. In the present study, the more informative finding is that the most consistent survival associations are observed in KIRC and STAD, where elevated TF expression remained independently linked to worse overall survival following adjustment for available clinical variables [38, 39]. At the same time, the modest diagnostic performance on ROC and the lack of a consistent stage-dependent trend in these cancers suggest that TF expression may better reflect underlying disease context and biological behavior than mere tumor burden. The immune analyses further support the view that TF-associated biology is cancer-type dependent. In KIRC, TF expression correlated positively with infiltrating plasmacytoid dendritic cells and effector memory T cells, whereas in STAD it showed negative associations with several immune populations, including CD8+ T cells, macrophages, and NK cells. These contrasting patterns imply that TF expression may be linked to different TME contexts rather than reflecting a pan-cancer immune phenotype [41]. Similarly, the observation that promoter methylation was increased in KIRC but not in STAD is consistent with the possibility that TF is regulated through multiple mechanisms in different tumor settings [40, 42]. It is important to emphasize that these analyses are inherently correlative and do not constitute evidence of direct immunomodulatory function [43]. Instead, these data provide a rationale for future mechanistic studies to determine whether TF is simply a marker of specific tumor states, or whether it plays a more direct role in shaping the tumor microenvironment [44]. The pharmacogenomic analyses revealed significant associations between TF expression and sensitivity to multiple compounds across CTRP and GDSC datasets, including agents related to proteasome, MAPK-associated, and epigenetic pathways. Because these analyses were based on cross-cell-line correlations, they should be regarded as prioritization signals rather than evidence of TF-dependent therapeutic vulnerability [45]. Nonetheless, the recurrence of certain drug classes across datasets suggests that TF-associated transcriptional states may coincide with recognizable treatment-response profiles and may be useful for generating future pharmacogenomic hypotheses [46]. To extend the analysis beyond cancer, we examined independent GEO datasets representing NASH, HF, and LC. In these cohorts, TF expression was consistently elevated relative to controls, and pathway analyses highlighted recurrent enrichment in extracellular matrix organization, metabolic reprogramming, mitochondrial function, and stress-responsive programs. Some of these pathway categories exhibited partial overlap with those identified in our analyses [47, 48]. We interpret this overlap cautiously: it points to convergent transcriptional contexts, but does not establish a common TF-driven mechanism across malignant and non-malignant diseases. Such a restrained interpretation is more consistent with the weight of currently available evidence [49, 50]. The in vitro experiments provide preliminary biological support for our computational findings, while simultaneously clarifying the current limits of the study. In 786-O cells, enforced TF expression enhanced cell viability, DNA synthesis, colony formation, migration, and invasion, whereas TF knockdown showed the opposite trend. In OA/PA-treated hepatocytes, TF modulation was associated with altered lipid accumulation and lipid metabolism-related gene expression. In AC16 cardiomyocytes, TF overexpression was linked to reactive oxygen species accumulation, reduced mitochondrial membrane potential, and increased apoptosis-related cell death. Collectively, these observations support the possibility that TF is related to proliferative, metabolic, and stress-related phenotypes across these selected cellular contexts. However, given that these experiments were conducted in vitro and lack rescue-based mechanistic dissection, they should be interpreted as supportive rather than definitive evidence. Overall, the present work should be viewed as an integrative and hypothesis-generating study. It does not establish TF as a universal oncogenic driver, nor as a unifying molecular bridge between cancer and cardio-metabolic disorders. Rather, this study identifies specific settings in which TF appears to be most relevant, particularly KIRC and STAD, and provides a structured basis for future work using single-cell profiling, functional rescue experiments, and in vivo models. Limitation Several limitations should be acknowledged. First, most analyses were based on bulk transcriptomic and proteomic datasets, which do not resolve cell-type-specific expression or spatial heterogeneity within the tumor microenvironment. Second, many of the key findings, including immune infiltration, drug-response correlations, and cross-disease pathway overlap, are association based and should not be interpreted as causal. Third, although in vitro validation was included, the functional experiments were limited to selected cell models and did not incorporate rescue strategies or direct interrogation of iron handling, ferroptosis, or oxidative stress mechanisms. Finally, the retrospective nature of public datasets introduces unavoidable constraints, including cohort heterogeneity, incomplete treatment and comorbidity annotations, and potential population bias, all of which may affect generalizability. Conclusion This integrative pan-cancer analysis offers a systematic characterization of TF expression, prognostic relevance, immune correlates, and pathway features across multiple tumor types. The findings support TF as a context-dependent biomarker candidate, with the most consistent survival associations observed in KIRC and STAD. Moreover, TF expression correlated with cancer-type-specific immune infiltration patterns, selected pharmacogenomic profiles, and partial pathway overlap with NASH, HF, and LC datasets. In vitro experiments conducted in hepatocyte, KIRC, and cardiomyocyte models provided preliminary biological support for these observations. Overall, these findings indicate that TF warrants further investigation as a biologically relevant factor in selected cancers and metabolic stress-associated conditions, while future single-cell, mechanistic, and in vivo studies will be required to define its precise functional role. Supplementary Information Supplementary Material 1. Supplementary Material 1. Supplementary Material 2. Supplementary Material 2.