Sage Journals: Discover world-class research

Abstract

Long non-coding RNAs (lncRNAs) can function as competing endogenous RNAs (ceRNAs) that rewire post-transcriptional regulation in glioblastoma (GBM). Previous GBM studies have focused on either single lncRNA ceRNA axis in isolation or used in silico predictions with small patient cohorts (< 200). In this study, we integrated RNA-seq data from 372 TCGA-GBM tumors, 5 matched adjacent TCGA-normal brain and 2 931 GTEx-normal brain (n = 3 308) samples to build an experimentally informed ceRNA atlas. Limma-voom differential analysis, intersection with 2 experimentally supported interaction databases (ENCORI and miRTarBase) distilled 517 high-confidence lncRNA–miRNA–mRNA triplets. Twelve hub lncRNAs coordinated 3 downregulated miRNAs and 262 target mRNAs enriched for cell-cycle, p53 signaling and homologous recombination pathways. Two co-expressed hubs, CYTOR and MIR4435-2HG, were significantly over-expressed in GBM tumors in comparison with normal brain tissue and independently predicted poor overall survival (log-rank P < .01). Their shared 25 targets include oncogenic YBX1, MDM4 and TGFBR1 mRNAs, underscoring the redundant regulation of oncogenic pathways, suggesting the need to explore combination lncRNA inhibition strategies. This population-scale analysis prioritizes CYTOR and MIR4435-2HG for functional interrogation and offers a framework for exploring biomarkers and RNA-targeted strategies in GBM.

Keywords

Glioblastoma (GBM)long non-coding RNA (lncRNA)competing endogenous RNA (ceRNA) network RNA-seq integrative analysis prognostic biomarkers

Introduction

Glioblastoma multiforme (GBM) remains one of the most aggressive and lethal forms of brain tumors, with a median survival of approximately 12-15 months despite improvements in surgical resection, radiotherapy, and chemotherapy.^1-3 Poor prognosis can be attributed to the heterogeneity, invasive growth patterns, and resistance to conventional therapies.^1-3 GBM’s diversity and rapid resistance to existing treatments mean many patients stop responding, so finding new therapeutic targets is essential.² In recent years, noncoding RNAs, especially long non-coding RNAs (lncRNAs), have emerged as pivotal regulators of gene expression in GBM, involved in complex post-transcriptional networks that regulate various hallmarks of cancer, including cell division, angiogenesis, and invasion.^4-6 Recent studies have shown that lncRNAs can act via the competing endogenous RNA (ceRNA) mechanism, in which RNA transcripts containing shared microRNA response elements (MREs) can compete for miRNA binding, thereby modulating gene stability and translation.^7-9 By acting as molecular sponges, lncRNAs, such as HOTAIR, MALAT1, and NEAT1, have been shown to sequester tumor-suppressive miRNAs, leading to derepression of key drivers such as FGF1, SOX2, and CD73.^10-12

Despite these advances, most studies focused on individual ceRNA axis in isolation in GBM, either in cell lines (A172 and U87-MG) or small patient cohorts, failing to integrate experimental interaction data.^13-15 In this study, we systemically constructed and validated a comprehensive GBM-specific ceRNA network by integrating large-scale RNA-seq data from 372 TCGA-GBM patients (primary tumors), 5 adjacent normal brain tissues of the TCGA-cohort, and 2931 normal brain samples from GTEx.^16,17 By leveraging experimentally validated interaction databases (ENCORI for lncRNA-miRNA pairs and miRTarBase for miRNA-mRNA pairs) and applying stringent correlation and differential-expression workflows, we pinpointed critical lncRNA-miRNA-mRNA triplets.^18,19 To further prioritize candidates, we overlaid clinical associations, examining expression versus patient survival, for our top lncRNAs, thereby highlighting those that likely drive GBM progression. This integrative approach not only enhances power and robustness over previous GBM ceRNA studies but also delivers a prioritized shortlist of functional ceRNA axes for future mechanistic and therapeutic exploration.

Methods

Data acquisition and processing

Raw RNA sequencing gene expression counts for GBM and normal brain tissue were obtained from TCGA and the GTEx projects. TCGA-GBM data (Primary Tumor n = 372, Solid Tissue Normal n = 5) were retrieved using the TCGAbiolinks package (v.2.34.1). Clinical metadata were retrieved in the same query. To maximize statistical power, GTEx-normal brain RNA sequencing data were accessed for 2931 normal brain samples spanning all 13 GTEx brain subregions via the recount3 package (project “BRAIN”). For both data sets, Ensembl gene IDs were harmonized by removing version numbers, and only genes present in all the data sets were retained for downstream analysis. GENCODE v38 (GRCh38.p13) gene annotation was downloaded and parsed using the R package, rtracklayer, to extract biotype and external gene names, enabling classification of transcripts as protein-coding mRNAs or lncRNAs.

Data normalization and filtering

For gene expression, raw counts were combined across TCGA-GBM tumor, TCGA-, and GTEx-normal brain samples, specifically utilizing STAR-count data. The dimensions of the matrix were 57 562 genes × 3308 samples. An edgeR DGEList object was created, supplying both counts and library information for filtering low-expressed genes, Trimmed Mean of M-values (TMM) normalization, and batch-corrected quality assessments prior to limma modeling. The combined matrix was filtered to retain genes expressed as counts per million (CPM) > 1. Normalization and library size correction were performed using the TMM method with the edgeR package. Voom precision weights were calculated in limma (v3.60.2). For miRNA expression, TCGA-GBM miRNA sequencing data (read counts) were normalized using the same TMM approach. To address potential technical variation arising from different sequencing platforms and tissue preservation methods between TCGA and GTEx data sets, we incorporated batch as a covariate in our linear model. The design matrix explicitly included both biological group (Tumor vs Normal) and technical batch (TCGA vs GTEx): design = model.matrix(~ group + batch). This approach estimates differential expression while statistically adjusting for systematic differences between data sets, following established best practices for cross-cohort RNA-seq integration.²⁰

Differential expression analysis

Differential gene and miRNA expression analyses were performed using the limma-voom pipeline. For genes (lncRNAs and miRNAs), a design matrix was fitted with lmfit followed by empirical Bayes moderation (eBayes) to compare GBM tumors versus all normal brain samples, accounting for batch effects. The batch covariate (TCGA vs GTEx) was included in the linear model to statistically adjust for platform-specific technical variation while preserving biological differences. Differentially expressed genes (DEGs) were defined as log₂ fold-change (FC) > 1 with Benjamini-Hochberg false discovery rate (FDR) < 0.05. miRNA differential analysis employed an analogous pipeline (miRNA_group factor: Tumor vs Normal).

Integration of experimentally validated interactions

ENCORI (StarBase v3.0, “CLIP-replicates across ⩾ 10 experiments”) database with high-confidence CLIP-supported human lncRNA-miRNA interaction pairs were downloaded. LncRNA gene names were mapped to Ensembl IDs using GENCODE annotation. Mature miRNA names were converted to precursor IDs via miRbaseConverter for harmonization with expression data. miRTarBase v9.0 (support type = reporter assay / western blot) was used to obtain validated miRNA-mRNA pairs. Human-specific interactions were retained, and gene names were mapped to Ensembl gene IDs. Mature miRNA names were also converted to precursor IDs for compatibility.

Construction of the GBM-specific ceRNA network

Differentially expressed (DE) lncRNAs (upregulated), DE miRNAs (downregulated) and DE mRNAs (upregulated) were intersected with the validated interaction sets to generate the ceRNA network—“up-lncRNA → down-miRNA → up-mRNA” triplets.

Survival analysis

Overall survival (OS) data for 372 patients with complete follow-up were extracted from TCGA data sets. For each identified hub lncRNA, a Kaplan-Meier curve was generated using median expression (survival v3.6-4, survminer v0.4.9); significance was assessed by 2-sided log-rank test (P < .05).

Expression analysis

Normalized log₂ CPM expression values for hub lncRNAs were calculated from raw counts in TCGA-GBM tumor, TCGA-normal brain, and GTEx-normal brain sample data. Expression differences between groups were visualized using boxplots (R, ggplot2), and statistical significance was tested with Wilcoxon rank-sum and Kruskal–Wallis tests. Significance was set at P < .05.

Functional enrichment analysis

Entrez IDs of each hub-specific mRNA set were submitted to clusterProfiler (v4.10.0). Gene Ontology Biological Process (GO-BP) and KEGG pathways were considered significant at q < 0.05. Dot-plots and bar-plots of the top terms were exported as high-resolution JPEGs.

Statistical analysis and software environment

All computations were performed in R 4.4.0 on macOS 14.4 (Apple M2 Pro, 32 GB RAM). Figures were created using ggplot2 v3.5.1 and ggpubr v0.6.0.

Results

Data acquisition, normalization and quality control for downstream analysis

TCGA-GBM tumor (Primary Tumor = 372 and Adjacent Normal = 5) and GTEx data set (n = 2931 samples) were integrated, retaining 57 562 genes for transcriptomic analysis. In addition, 1881 miRNAs were used from TCGA-GBM data set (Primary Tumor = 240 and Adjacent Normal = 37) for downstream analysis. The counts for each of these genes were normalized using the edgeR-voom pipeline. To ensure robust and comparable downstream analyses, we assessed the quality and distribution of RNA-seq data following harmonization and batch correction of 3308 samples spanning 372 TCGA-GBM tumors, 5 TCGA-normal brain, and 2,931 GTEx-normal brain tissues (Figure 1A and B). As shown in Figure 1A, the sample-wise median log₂-CPM values and interquartile range (IQR) were observed for mRNA/lncRNA and miRNAs between normal and brain tissues (Figure 1A to D), depicting that the distribution of log₂-CPM is uniform between the different sources. A summary of the combined and harmonized data sets used for subsequent analyses is provided in Figure 1E. After filtering, a total of 57 562 genes were quantified across 3308 samples (372 TCGA-GBM tumor, 5 TCGA-normal brain and 2931 GTEx-normal brain), while 1881 miRNAs were quantified in 277 samples (240 TCGA-GBM tumor, 37 TCGA-normal brain). This indicates successful harmonization and batch correction across heterogeneous sources, resulting in comparable global expression profiles and minimized technical variation.

Figure 1.

Assessment of RNA-seq data quality and sample distribution across TCGA-GBM and normal brain cohorts. (A) Boxplot showing the distribution of sample-wise median log₂ CPM values for mRNA/lncRNA across TCGA-GBM tumor samples, TCGA-adjacent normal brain, and GTEx-normal brain tissues. Median and interquartile range (IQR) are annotated for each group. (B) Boxplot summarizing the sample-wise IQR of log₂ CPM for each group, reflecting global expression variability. (C) Sample-wise median log₂ CPM for miRNAs in TCGA-normal and tumor samples. (D) Boxplot of sample-wise IQR of log₂ CPM for miRNAs in TCGA-normal and tumor groups. (E) Summary table of the harmonized and combined data sets, indicating the number of genes and samples for each RNA biotype and subset used in downstream differential expression analyses.

Differential expression analysis of genes between different databases

Building on these harmonized data, we applied a stepwise ceRNA workflow that combines differential expression with experimentally validated lncRNA-miRNA-mRNA interactions to pinpoint GBM-specific ceRNA regulatory triplets (Figure 2). DE analysis was performed using the limma pipeline to compare mRNAs, lncRNAs, and miRNAs between TCGA-GBM tumor and normal brain samples (TCGA and GTEx). Figure 3 shows heatmaps illustrating the distinct expression patterns of DE genes of the whole transcriptome (Figure 3A; gene-level] and significantly downregulated miRNAs across all samples (Figure 3B; miRNA-level]. While the gene-level heatmap demonstrates clear segregation between tumor and normal samples, reflecting extensive transcriptional dysregulation characteristic of GBM, the miRNA-level heatmap reveals greater heterogeneity among tumor samples, with certain tumors exhibiting expression patterns like normal. Such heterogeneity is more likely biological rather than technical, as global quality metrics (median log₂-CPM, IQR) confirmed comparable data quality across cohorts (Figure 1 C and D]. Differentially expressed lncRNA, mRNA and miRNA were visualized as volcano plots (Figure 4). In the mRNA and lncRNA panels (Figure 4A and B), both upregulated and downregulated transcripts were detected in GBM, indicating broad transcriptomic reprogramming in tumors. For lncRNA DE analysis, 516 lncRNAs were upregulated, whereas 433 lncRNAs were downregulated. mRNA DE analysis revealed that 2541 mRNAs were upregulated, whereas 3097 mRNAs were downregulated. In contrast, 4 miRNAs were found to be downregulated in tumor samples in comparison with normal brain samples (Figure 4D). These findings underscore the widespread molecular alterations in GBM and support further integrative analysis of both coding and non-coding RNA classes.

Figure 2.

Funnel overview of the GBM ceRNA workflow.

Figure 3.

Unsupervised clustering of normalized expression data in GBM and normal brain samples. (A) Gene-level heatmap: Hierarchical clustering of z-score normalized log2 CPM expression values for all differentially expressed mRNAs and lncRNAs across TCGA-GBM primary tumor samples (orange), TCGA adjacent normal brain samples (blue), and GTEx-normal brain samples (green). Rows represent genes, columns represent samples. Distinct segregation between tumor and normal samples reflects extensive transcriptional dysregulation in GBM, while the tight clustering of normal samples highlights their relative homogeneity. (B) miRNA-level heatmap: Hierarchical clustering of z-score normalized log2-CPM values for all significantly differentially expressed miRNAs in TCGA-GBM tumors (orange) and TCGA-normal brain samples (blue). Rows represent miRNAs, columns represent samples. Increased heterogeneity is observed among tumor samples, with a subset displaying miRNA expression patterns similar to normal brain tissue, consistent with the diverse molecular subtypes of GBM. Key DE miRNAs are annotated.

Figure 4.

Overview of differential expression analysis in TCGA-GBM versus normal brain (TCGA-normal brain and GTEx-normal brain). Volcano plot for (A) protein-coding genes, (B) lncRNAs, and (C) miRNAs, showing log₂ fold-change (x-axis) against log₁₀ adjusted P-value (y-axis). Each point represents a transcript; colors indicate upregulation (orange), downregulation (blue), or not significant (gray), based on thresholds of log₂ FC > 1 and FDR < 0.05. Top significantly dysregulated transcripts are labeled in each panel. (D) Summary table listing the total number of upregulated and downregulated transcripts detected for each RNA class using the same significance criteria as panels A-C.

The diagram tracks each major processing step from raw multi-cohort RNA-seq input (top) through normalization, differential-expression filtering, and experimental-validation layers to the final set of validated lncRNA-miRNA-mRNA triplets and prognostic hub lncRNAs (bottom).

ceRNA triplet (lncRNA -| miRNA -> mRNA) construction

After the differentially expressed genes were identified, upregulated mRNAs (Figure 5A] were overlapped with miRTarBase-validated mRNAs, yielding 1984 genes (13.1%) present in both sets. Similarly, among lncRNAs, 96 genes (9.7%) were shared between the set of upregulated lncRNAs and those supported by ENCORI (Figure 5B]. For miRNAs, only 3 downregulated miRNAs were identified as common to all 3 groups: differentially expressed, ENCORI-validated, and miRTarBase-validated sets (Figure 5 C]. The pipeline summary highlights each filtering step: from the initial pool of differentially expressed genes, through validation with ENCORI and miRTarBase, to the final intersection, resulting in a network of 12 lncRNAs, 3 miRNAs, and 262 mRNAs that assembled into 517 unique ceRNA triplets (Figure 5D]. This integrative approach maximizes both the biological relevance and experimental support of the inferred ceRNA interactions in GBM.

Figure 5.

Construction of the GBM ceRNA network using integrated differential expression and experimentally validated interactions. (A) Venn diagram showing overlap between upregulated mRNAs in GBM and experimentally supported mRNA targets from miRTarBase. (B) Overlap between upregulated lncRNAs and ENCORI-validated lncRNAs. (C) Venn diagram illustrating the intersection of downregulated miRNAs, ENCORI-supported miRNAs, and miRTarBase-validated miRNAs. (D) Summary table of the ceRNA network construction pipeline, including the number of lncRNAs, miRNAs, and mRNAs at each filtering step and the total number of final ceRNA triplets.

Pathway enrichment analysis using GO and KEGG databases

To gain functional insights into the biological relevance of mRNAs regulated through the GBM ceRNA network, we performed pathway enrichment analysis using both GO and KEGG databases. GO biological process analysis revealed a strong enrichment for cell cycle-associated terms, including regulation of cell cycle phase transition, chromosome segregation, mitotic nuclear division, and DNA-templated replication (Figure 6A). These processes are consistent with the hyperproliferative nature of GBM and reflect the central role of post-transcriptional regulation in maintaining mitotic control. KEGG pathway enrichment further supported these findings, with top-ranked pathways including cell cycle, homologous recombination, proteoglycans in cancer, and the p53 signaling pathway (Figure 6B). These pathways underscore the involvement of ceRNA-regulated mRNAs in key tumorigenic processes such as genomic instability, extracellular matrix remodeling, and evasion of cell cycle checkpoints. Together, these data suggest that the ceRNA network plays a functional role in promoting GBM malignancy by modulating post-transcriptional regulation of genes critical to cell division and cancer progression.

Figure 6.

Pathway enrichment analysis of mRNAs within the GBM ceRNA network. (A) GO-biological process enrichment of ceRNA-associated mRNAs. The top 15 significantly enriched terms are shown, ranked by adjusted P-value. Bar color reflects the gene ratio (ie, the proportion of mRNAs involved in each process). Prominent processes include cell cycle phase transition, chromosome segregation, and mitotic nuclear division, reflecting active proliferation and genomic instability in GBM. (B) KEGG pathway enrichment analysis of the same mRNAs. The top enriched pathways include cell cycle, homologous recombination, proteoglycans in cancer, and the p53 signaling pathway, indicating the involvement of ceRNA-regulated mRNAs in tumor progression, DNA damage repair, and cell cycle dysregulation.

Expression and survival analysis of ceRNA hub lncRNAs

We further evaluated the prognostic relevance of the lncRNAs involved in the identified ceRNA network using expression and survival analysis (Figure 7 and 8). For each lncRNA, expression (log₂ CPM) was compared across TCGA-GBM tumor samples (red), TCGA-normal brain samples (blue), and GTEx-normal brain samples (green) (Figure 7). Boxplots demonstrate that lncRNAs exhibited highly significant differential expression between tumor and normal tissues. Seven out of 12 lncRNAs were markedly upregulated in tumor samples compared to both TCGA- and GTEx-normal brain tissue. This was followed by Kaplan-Meier survival analysis on 12 hub lncRNAs using TCGA-GBM primary tumor data (Figures 8 and 9). The overall survival between high- and low-expression groups was compared using log-rank tests (P-value < 0.05). Among the tested lncRNAs, MIR4435-2HG (P = .0024) and CYTOR (P = .0032) showed statistically significant differences in survival, with high expression levels associated with poorer prognosis. Notably, these lncRNAs also exhibited a significantly higher expression in GBM samples in comparison with normal tissues, warranting further functional validation and clinical exploration (Figure 8).

Figure 7.

Expression profiles of representative ceRNA network lncRNAs across sample groups.

Figure 8.

Kaplan-Meier survival curves for hub lncRNAs identified in the GBM ceRNA network.

Figure 9.

Univariate Cox regression analysis of hub lncRNAs associated with overall survival in TCGA-GBM.

Shared mRNA regulation between hub lncRNAs (CYTOR and MIR4435-2HG) with a prognostic potential

The constructed ceRNA network revealed a notable set of mRNAs that are co-regulated by CYTOR and MIR4435-2HG (Figure 10A and Table 1). These mRNAs are targeted by the same miRNA that interact with both lncRNAs based on the experimentally validated ENCORI and mirTarBase databases, suggesting a shared regulatory axis. Functional annotation of these shared mRNA targets indicates convergence on key oncogenic pathways relevant to GBM, including cell proliferation, immune modulation, invasion, and angiogenesis. Several of these mRNAs, such as YBX1, TGFBR1, HOXA10, and CD44, have previously been implicated in GBM pathogenesis and are associated with poor patient prognosis (Table 1). STRING protein-protein interaction analysis of the 25 shared mRNA targets identified 7 experimentally supported interactions (average confidence score: 0.636), revealing functionally coherent modules centered on key oncogenic drivers. CD44 emerged as the most connected hub (3 interactions), showing strong interactions with TGFBR1 (score = 0.849) and THBS1 (score = 0.767), both critical for TGF-β signaling and extracellular matrix remodeling in GBM. A second module linked translation-related proteins METTL1, RPL7, and YBX1 (score = 0.836). While the network shows moderate overall connectivity, the clustering of interactions around established GBM drivers—including CD44 (stemness), YBX1 (mTOR/MAPK signaling), and TGFBR1 (invasion)—supports functional convergence of ceRNA-regulated targets (Figure 10B]. The dual regulation by CYTOR and MIR4435-2HG highlights these mRNAs as potential network hubs and underscores their importance as effectors of lncRNA-mediated gene regulation in GBM.

Figure 10.

Focused ceRNA and protein-protein interaction networks for CYTOR and MIR4435-2HG in GBM. (A) The ceRNA subnetwork centers on CYTOR and MIR4435-2HG, which both interact with hsa-miR-379 targeting multiple oncogenic mRNAs. Blue, orange, and red-orange nodes represent lncRNAs, the miRNA, and mRNAs, respectively. Node size indicates biological relevance, highlighting GBM-associated targets such as YBX1, MDM4, CD44, and TGFBR1. Arrows denote regulatory direction (lncRNA → miRNA → mRNA). (B) PPI network derived from panel (A) showing hub proteins (⩾ 2 connections; red) and other interactors (blue). Edge thickness reflects interaction strength. Key hubs—TGFBR1, CD44, YBX1, THBS1, and METTL1- suggest convergence of the ceRNA axis on central oncogenic regulators in GBM.

Table 1.

Shared mRNA targets of CYTOR and MIR4435-2HG within the GBM ceRNA network.

Symbol	Gene name (short)	Role in GBM
YBX1	Y-box binding protein 1	YBX1 promotes GBM growth by enhancing mTOR signaling through translational upregulation of CCT4 and autoregulation of its own expression.²¹
MDM4	MDM4 regulator of p53	Negative regulator of p53; often overexpressed in GBM, contributing to p53 pathway inactivation; high levels may correlate with worse outcome.²²
ABHD5	Abhydrolase domain containing 5	Up-regulation of ABHD5 augments ATGL-mediated lipolysis, increasing lipid-droplet turnover and supporting the metabolic requirements of GBM cells.²³
MCC	Regulator of WNT signaling pathway	Up-regulated MCC correlates with poorer prognosis in GBM, and experimental knock-down accelerates GBM cell proliferation and migration, highlighting its role as a regulator of tumor growth.²⁴
SPRY4	Sprouty RTK signaling antagonist 4	Up-regulated lncRNA SPRY4 functions oncogenically in glioma, accelerating proliferation, invasion and angiogenesis through miR-101-3p → EZH2 / VEGFA signaling.²⁵
RNF144B	Ring finger protein 144B	Elevated RNF144B in tumor-associated endothelial cells supports pro-angiogenic signaling, sustaining vascularization and growth of GBM.²⁶
SLC16A10	Solute carrier family 16 member 10	High SLC16A10 levels in GBM endothelial and tumor cells boost aromatic-amino-acid transport, support tumor metabolism, and are linked to poorer patient survival.^27,28
HOXA10	Homeobox A10	HOXA10 enhances homologous-recombination DNA-repair pathways in GBM cells, thereby reducing their susceptibility to temozolomide.²⁹
RPL7	Ribosomal protein L7	Up-regulated RPL7 boosts ribosome biogenesis and translational output in GBM, matching the tumor’s high proliferative demand and emerging as a non-invasive urinary-EV biomarker.³⁰
FZD6	Frizzled class receptor 6	FZD6 overexpression drives a mesenchymal, therapy-resistant GBM phenotype by activating CaMKII-TAK1-NFκB/STAT3 signaling while dampening canonical Wnt activity.³¹
PTAR1	Prenyltransferase alpha repeat 1	PTAR1, a predicted Rab-geranylgeranyl-transferase subunit, is co-expressed with the EMT driver ZEB1 in GBM and may promote vesicle-trafficking–dependent adhesion and invasion of tumor cells.³²
TGFBR1	TGF-beta receptor 1	In GBM cell models, TGFBR1 mediates Tim-1/miR-133a-dependent control of proliferation, invasion, and Wnt/β-catenin signaling, marking it as a critical effector of GBM malignancy.³³
ADM	Adrenomedullin	In GBM, AM is upregulated in hypoxic tumor regions, where it acts as an autocrine/paracrine angiogenic survival factor via CRLR/RAMP2/3 signaling, promoting tumor growth and neovascularization.³⁴
CD44	CD44 molecule	High CD44 expression identifies invasive, therapy-resistant mesenchymal GBM cells and its abundance independently predicts poor patient survival.^35-37
EPS8	EGFR pathway substrate 8	EPS8 is upregulated in GBM, activates PI3 K/AKT signaling, drives cell proliferation and invasion, and is linked to poorer patient survival.³⁸
METTL1	Methyltransferase 1 (tRNA)	RNA methylation enzyme; new research links tRNA methylation with cancer, little direct evidence of role in GBM.^39,40
RAP1B	Ras-related protein RAP1B	Up-regulated RAP1B drives cytoskeleton-dependent proliferation and invasion in GBM, and its high expression marks poorer patient survival.^41,42
BEST3	Bestrophin 3	BEST3 encodes a Ca²⁺-activated chloride channel; in GBM these channels facilitate cell-volume regulation and migration, processes that support rapid invasion and therapy resistance.⁴³
CSRP2	Cysteine and glycine rich protein 2	CSRP2 is over-expressed in GBM, where it amplifies Notch signaling to fuel tumor cell proliferation, invasion and stem-like traits, and its high expression portends poorer patient survival.⁴⁴
SERPINA1	Serpin family A member 1 (alpha-1 antitrypsin)	SERPINA1 is over-expressed in GBM, where it drives tumor-cell proliferation and its high levels predict shorter patient survival.⁴⁵
THBS1	Thrombospondin 1	Up-regulated THBS1 increases GBM cell invasion and angiogenesis via CD47 signaling, and its high expression marks the mesenchymal subtype and poorer patient survival.^46,47
SOCS1	Suppressor of cytokine signaling 1	SOCS1 is over-expressed in IDH-wild-type GBM, and its high levels independently predict poor survival, underscoring this JAK/STAT brake as a potential immunotherapy target.^48,49
CD226	CD226 molecule	Impaired CD226 signaling on tumor-infiltrating lymphocytes weakens cytotoxic immunity in GBM, while re-engaging the CD226/CD155/CD112 axis restores potent γδ-T-cell-mediated tumor lysis.^50,51
ZNF460	Zinc finger protein 460	Bioinformatic screens in PTEN-null GBM highlighted ZNF460 as a top candidate transcription factor upstream of the Galectin-9 immune-checkpoint axis, positioning it as a putative pro-tumor regulatory hub in GBM.⁵²
IFNLR1	Interferon-λ receptor 1	Evidence does not support an oncogenic role for IFN-λ receptor 1 in GBM.

Co-expression analysis between CYTOR and MIR4435-2HG in TCGA-GBM cohort

As both the lncRNAs regulated the same mRNAs and, therefore, may share a regulatory axis, we conducted a correlation analysis between these lncRNAs in TCGA-GBM tumor samples to further probe their relationship. A Pearson correlation analysis revealed a strong positive association between CYTOR and MIR4435-2HG expression (r = .846, R² = 0.716, P < 2e−16; Figure 11). Next, the genomic location of these two lncRNAs was mapped to chromosome 2, with the exact coordinates of CYTOR being chr2:87454781-87630751 and MIR4435-2HG being chr2:111006015-111495070. This high degree of co-expression and proximity on chromosome 2 suggests potential co-regulation or shared upstream regulatory mechanisms and supports their cooperative involvement within the ceRNA network in GBM.

Figure 11.

Correlation between CYTOR and MIR4435-2HG expression in GBM tumor samples.

Discussion

Our integrative analysis unites the largest publicly available GBM and normal-brain RNA-seq compendium to date with 2 experimentally supported interaction atlases, revealing a GBM-specific ceRNA landscape that is both experimentally anchored and clinically annotated. Consistent with the poor outcomes that still characterize GBM even after temozolomide-based chemoradiation, we found widespread transcriptomic rewiring across mRNAs, miRNAs, and lncRNAs.¹ The uniform global quality control metrics across TCGA and GTEx samples argue that these differences are overwhelmingly biological rather than technical, providing data for robust downstream network construction.

While our ceRNA network construction utilized 2 gold-standard databases (miRTarBase 2025 and ENCORI) with stringent experimental validation criteria, we acknowledge that cross-validation with additional databases such as LncBase or DIANA-LncBase could provide complementary evidence.^19,53,54 However, DIANA-LncBase v3 (released 2019/2020) has not been updated in over 5 years, while our selected databases provide the most current experimentally validated interactions, with ENCORI requiring ⩾ 10 independent CLIP-seq experiments. This stringency ensures high confidence ceRNA interactions, and future validation using orthogonal databases or independent patient cohorts would further strengthen these findings.

Most published GBM ceRNA studies focus on a single lncRNA-miRNA-mRNA axis and are limited to small patient cohorts often under 200 tumors, with key findings based on cell lines or in silico predictions. This highlights the need for, and the strength of, a transcriptome-wide ceRNA approach grounded in experimental evidence, as implemented in our pipeline.^12,14,15 We executed the pipeline at a population scale: 3308 brain specimens, stringent DE filters, and dual-database interaction validation (ENCORI and miRTarBase).^19,54 This strategy funneled > 57 000 measured genes down to a network of 517 triplets linking 12 hub lncRNAs to 262 mRNAs through 3 repressed miRNAs. GO- and KEGG-pathway analysis revealed that mRNAs present in the ceRNA network played a role in cell cycle,- DNA-replication and p53-signalling terms, complementing the aggressiveness, unchecked proliferation and genomic instability of GBM.³ Enrichment of homologous-recombination genes also supports emerging evidence that GBM cells exploit DNA-repair pathways to survive genotoxic stress, potentially explaining the limited durability of alkylating chemotherapy.³ Overall, our study distills a large, well-validated data set into a concise list of lncRNA-centered ceRNA interactions that are most relevant to GBM biology and therapy development.

Among the twelve lncRNA hubs, CYTOR and MIR4435-2HG were chosen for further analysis for 4 reasons: (1) marked over-expression in tumors versus 2 normal-brain cohorts, (2) strong effect on prognosis as suggested by the survival curves, (3) co-expression (Pearson r = .85) in GBM samples, and (4) they reside ~23 Mb apart on chromosome 2 suggesting a similar upstream mechanism. Their shared repression of 3 miRNAs funnels de-repression to 25 convergent mRNA targets, including YBX1, MDM4 and TGFBR1, each with established oncogenic functions in GBM. For example, YBX1 drives mTORC1/2 activation and GBM growth,²¹ while MDM4 amplification disables p53 checkpoint activity and portends poor outcome.²² The dual-lncRNA control of these nodes suggests a “redundant safeguard” whereby loss of one sponge may be buffered by the other, suggesting the need to explore possible combinatorial lncRNA inhibition strategies. Beyond cell cycle- genes, the network captured transcripts that remodel tumor metabolism and the niche. ABHD5, a co-activator of ATGL, links lipid-droplet catabolism to energy homeostasis and has recently been tied to GBM growth under nutrient stress.²³ SPRY4, another hub-controlled lncRNA, promotes VEGFA-driven angiogenesis through a miR-101-3p/EZH2 axis,²⁵ while RNF144B although classically viewed as a p53-induced tumor suppressor has been reported to regulate endothelial-cell signaling and could, when over-expressed in the GBM vasculature, promote angiogenesis.²⁶ Together, these findings indicate that ceRNA crosstalk intersects not only intrinsic tumor-cell programs but also metabolic adaptation and pro-angiogenic cues, reinforcing that lncRNA-targeted therapies might exert multi-pronged anti-tumor effects.

The strong co-expression between CYTOR and MIR4435-2HG (r = .85, P < 2 × 10-)¹⁶ combined with their location on the same chromosome (~23 Mb apart on chromosome 2) suggests they may be regulated together. To test whether their proximity drives coordinated expression, future studies could use chromatin conformation data (Hi-C) to see if they occupy the same regulatory neighborhoods (topologically associating domains), or ChIP-seq to identify shared enhancers or transcription factor binding sites. This would clarify whether their co-expression results from shared local regulatory elements or from responding independently to the same cancer signals. While the data sets needed for these analyses are currently limited in GBM, understanding how these 2 lncRNAs coordinate could reveal new regulatory mechanisms in glioblastoma.

Prognostic modeling underscored CYTOR and MIR4435-2HG as robust survival markers, but 5 additional hubs trended toward poorer outcome, warranting validation in prospective cohorts and functional assays. Limitations include reliance on bulk RNA seq- (unable to resolve cell-type-specific ceRNA wiring) and a lack of experimental perturbation to confirm directionality of repression. Future work should integrate single-cell or spatial transcriptomics to localize ceRNA interactions, employ CRISPR interference, siRNA or ASO knockdown to test lncRNA causality, and explore synthetic-lethal combinations that jointly disrupt multiple sponge nodes.

Beyond their roles in proliferation and survival, CYTOR and MIR4435-2HG may influence the immune microenvironment of GBM, as shown by their modulation of TGFBR1, CD44, SOCS1, and CD226. Although the central nervous system was once considered immune-privileged, recent evidence shows that GBMs are highly infiltrated by microglia and bone-marrow-derived macrophages that sustain an immunosuppressive milieu.⁵⁵ MIR4435-2HG exhibits context-dependent immune regulation: in colorectal cancer it acts as a tumor suppressor within the immune stroma, where its loss increases polymorphonuclear myeloid-derived suppressor cells and impairs fatty-acid metabolism, whereas in hepatocellular and gastric cancers it promotes M2 macrophage polarization and immune-checkpoint up-regulation (CTLA4, PDCD1), fostering immune evasion.^56,57 Similarly, CYTOR has been shown to correlate with immune-cell infiltration, cytokine signaling, and checkpoint expression across multiple tumor types, including melanoma, where its expression was associated with improved response to anti-PD-1 therapy and enrichment of immune-related pathways.⁵⁸ These observations suggest that both lncRNAs could contribute to macrophage-mediated immunosuppression or checkpoint-linked immune escape in GBM. Future studies integrating immune-cell deconvolution, single-cell RNA-seq, or spatial transcriptomics will be instrumental in clarifying their roles in shaping the GBM immune landscape.

In conclusion, our population scale—experimental database-anchored ceRNA atlas nominates a concise panel of clinically actionable lncRNAs, headed by CYTOR and MIR4435-2HG, which, through pathway analysis, suggest an effect on proliferative, metabolic, and angiogenic programs in GBM. These hubs, together with their shared miRNA and mRNA partners, form a rational starting point for RNA-targeted therapeutic development and biomarker-driven patient stratification.

Footnotes

Acknowledgements

The authors thank Dr Jessie King for her assistance in proofreading the manuscript.

ORCID iD

Zohaib Rana

Ethical Considerations

Not applicable; study uses only publicly available human RNA-seq data sets (TCGA, GTEx).

Author Contributions

Zohaib Rana: Conceptualization; Investigation; Writing—original draft; Methodology; Validation; Visualization; Writing—review & editing; Software; Formal analysis; Project administration; Data curation.

Joke Grans: Visualization; Writing—review & editing.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data Availability Statement

All processed data and analysis scripts are available on github:

References

Stupp

Mason

Van Den Bent

, et al. Radiotherapy plus concomitant and adjuvant temozolomide for glioblastoma. N Engl J Med. 2005;352:987-996. doi:10.1056/nejmoa043330

Pouyan

Ghorbanlo

Eslami

, et al. Glioblastoma multiforme: insights into pathogenesis, key signaling pathways, and therapeutic strategies. Mol Cancer. 2025;24:58. doi:10.1186/s12943-025-02267-0

Singh

Dey

Barik

, et al. Glioblastoma at the crossroads: current understanding and future therapeutic horizons. Signal Transduct Target Ther. 2025;10:213. doi:10.1038/s41392-025-02299-4

Yadav

Pal

Rubstov

, et al. LncRNAs associated with glioblastoma: from transcriptional noise to novel regulators with a promising role in therapeutics. Mol Ther—Nucleic Acids. 2021;24:728-742. doi:10.1016/j.omtn.2021.03.018

Stackhouse

Gillespie

Willey

CD.

Exploring the roles of lncRNAs in GBM pathophysiology and their therapeutic potential. Cells. 2020;9:2369. doi:10.3390/cells9112369

Hazra

Debnath

Tuppad

Glioblastoma stem cell long non-coding RNAs: therapeutic perspectives and opportunities. Front Genet. 2024;15:1416772. doi:10.3389/fgene.2024.1416772

Yang

, et al. Study on the Relationship between the miRNA-centered ceRNA Regulatory Network and Fatigue. J Mol Neurosci. 2021;71:1967-1974. doi:10.1007/s12031-021-01845-3

Salmena

Poliseno

Tay

Kats

Pandolfi

PP.

A ceRNA hypothesis: the rosetta stone of a hidden RNA language?

Cell. 2011;146:353-358. doi:10.1016/j.cell.2011.07.014

Liu

Xue

, et al. Competitive endogenous RNA is an intrinsic component of EMT regulatory circuits and modulates EMT. Nat Commun. 2019;10:1637. doi:10.1038/s41467-019-09649-1

10.

Lulli

Buccarelli

Ilari

, et al. Mir-370-3p Impairs glioblastoma stem-like cell malignancy regulating a complex interplay between HMGA2/HIF1A and the oncogenic long non-coding RNA (lncRNA) NEAT1. Int J Mol Sci. 2020;21:3610. doi:10.3390/ijms21103610

11.

Yao

long Zheng

, et al. Knockdown of long non-coding RNA HOTAIR inhibits malignant biological behaviors of human glioma cells via modulation of miR-326. Oncotarget. 2015;6:21934-21949. doi:10.18632/oncotarget.4290

12.

Xiong

Wang

Yuan

LncRNA MALAT1/miR-129 axis promotes glioma tumorigenesis by targeting SOX2. J Cell Mol Med. 2018;22:3929-3940. doi:10.1111/jcmm.13667

13.

Wang

Zhang

Zeng

Tan

CeRNA network analysis reveals prognostic markers for glioblastoma. Oncol Lett. 2019;17:5545-5557. doi:10.3892/ol.2019.10275

14.

Bazrgar

Mirmotalebisohi

Ahmadi

, et al. Comprehensive analysis of lncRNA-associated ceRNA network reveals novel potential prognostic regulatory axes in glioblastoma multiforme. J Cell Mol Med. 2024;28:e18392. doi:10.1111/jcmm.18392

15.

Jang

Yoon

Roh

Kim

LncRNA H19 acts as a ceRNA to promote glioblastoma malignancy by sponging miR-19b-3p and upregulating SERPINE1. Cancer Cell Int. 2025;25:217. doi:10.1186/s12935-025-03868-x

16.

The Cancer Genome Atlas Research Network. Comprehensive genomic characterization defines human glioblastoma genes and core pathways. Nature. 2008;455:1061-1068. doi:10.1038/nature07385

17.

The GTEx Consortium Aguet F, Anand

, et al. The GTEx Consortium atlas of genetic regulatory effects across human tissues. Science. 2020;369:1318-1330. doi:10.1126/science.aaz1776

18.

Chaudhary

Banerjee

Decoding the non-coding: tools and databases unveiling the hidden world of “Junk” RNAs for innovative therapeutic exploration. ACS Pharmacol Transl Sci. 2024;7:1901-1915. doi:10.1021/acsptsci.3c00388

19.

Cui

Huang

, et al. MiRTarBase 2025: updates to the collection of experimentally validated microRNA–target interactions. Nucleic Acids Res. 2025;53:D147-D156. doi:10.1093/nar/gkae1072

20.

Law

Chen

Shi

Smyth

GK.

Voom: precision weights unlock linear model analysis tools for RNA-seq read counts. Genome Biol. 2014;15:R29. doi:10.1186/gb-2014-15-2-r29

21.

Wang

Zhu

You

, et al. Upregulated YB-1 protein promotes glioblastoma growth through a YB-1/CCT4/mLST8/mTOR pathway. J Clin Invest. 2022;132:e146536. doi:10.1172/jci146536

22.

Furgason

Koncar

Michelhaugh

, et al. Whole genome sequence analysis links chromothripsis to EGFR, MDM2, MDM4, and CDK4 amplification in glioblastoma. Oncoscience. 2015;2:618-628. doi:10.18632/oncoscience.178

23.

Kou

Geng

Guo

Lipid metabolism in glioblastoma: from de novo synthesis to storage. Biomedicines. 2022;10:1943. doi:10.3390/biomedicines10081943

24.

Nguyen

Huang

Juang

, et al. The mutated in colorectal cancer (MCC) gene can serve as a potential biomarker of glioblastoma. Front Oncol. 2024;14:1435605. doi:10.3389/fonc.2024.1435605

25.

Wang

Chen

Wang

, et al. LncRNA SPRY4-IT1 facilitates cell proliferation and angiogenesis of glioma via the miR-101-3p/EZH2/VEGFA signaling axis. Cancer Med. 2023;12:7309-7326. doi:10.1002/cam4.5517

26.

DeSouza

Chen

Patel

Maher

Kim

AH.

Long, noncoding RNA dysregulation in glioblastoma. Cancers. 2021;13:1604. doi:10.3390/cancers13071604

27.

Bao

Xie

, et al. Protein expression and functional relevance of efflux and uptake drug transporters at the blood–brain barrier of human brain and glioblastoma. Clin Pharmacol Ther. 2020;107:1116-1127. doi:10.1002/cpt.1710

28.

Luo

, et al. Systematic and multi-omics prognostic analysis of lysine acetylation regulators in glioma. Front Mol Biosci. 2021;8:587516. doi:10.3389/fmolb.2021.587516

29.

Kim

Chung

Park

CK.

HOXA10 is associated with temozolomide resistance through regulation of the homologous recombinant DNA repair pathway in glioblastoma cell lines. Genes Cancer. 2014;5:165-174. doi:10.18632/genesandcancer.16

30.

Hallal

Tűzesi

Sida

, et al. Glioblastoma biomarkers in urinary extracellular vesicles reveal the potential for a “liquid gold” biopsy. Br J Cancer. 2024;130:836-851. doi:10.1038/s41416-023-02548-9

31.

Huang

Alvarez

Pangeni

, et al. A regulatory circuit of miR-125b/miR-20b and Wnt signalling controls glioblastoma phenotypes through FZD6-modulated pathways. Nat Commun. 2016;7:12885. doi:10.1038/ncomms12885

32.

Rosmaninho

Mükusch

Piscopo

, et al. Zeb1 potentiates genome-wide gene transcription with Lef1 to promote glioblastoma cell invasion. EMBO J. 2018;37:e97115. doi:10.15252/embj.201797115

33.

Wei

Peng

Shao

Zhou

Downregulation of Tim-1 inhibits the proliferation, migration and invasion of glioblastoma cells via the miR-133a/TGFBR1 axis and the restriction of Wnt/β-catenin pathway. Cancer Cell Int. 2021;21:347. doi:10.1186/s12935-021-02036-1

34.

Metellus

Voutsinos-Porche

Nanni-Metellus

, et al. Adrenomedullin expression and regulation in human glioblastoma, cultured human glioblastoma cell lines and pilocytic astrocytoma. Eur J Cancer. 2011;47:1727-1735. doi:10.1016/j.ejca.2011.02.021

35.

Inoue

Ohnishi

Nishikawa

, et al. A narrative review on CD44’s role in glioblastoma invasion, proliferation, and tumor recurrence. Cancers. 2023;15:4898. doi:10.3390/cancers15194898

36.

Kolliopoulos

Ali

Castillejo-Lopez

Heldin

CD44 depletion in glioblastoma cells suppresses growth and stemness and induces senescence. Cancers. 2022;14:3747. doi:10.3390/cancers14153747

37.

Ballestín

Armocida

Ribecco

Seano

Peritumoral brain zone in glioblastoma: biological, clinical and mechanical features. Front Immunol. 2024;15:1347877. doi:10.3389/fimmu.2024.1347877

38.

Yang

Guan

QL.

EPS8 is a potential oncogene in glioblastoma. Onco Targets Ther. 2019;12:10523-10534. doi:10.2147/ott.s227739

39.

Wang

Fang

, et al. METTL1-modulated LSM14A facilitates proliferation and migration in glioblastoma via the stabilization of DDX5. Iscience. 2024;27:110225. doi:10.1016/j.isci.2024.110225

40.

Yang

Wang

Huang

Prognostic role of METTL1 in glioma. Cancer Cell Int. 2021;21:633. doi:10.1186/s12935-021-02346-4

41.

She

Cui

, et al. MiR-128 and miR-149 enhance the chemosensitivity of temozolomide by Rap1B-mediated cytoskeletal remodeling in glioblastoma. Oncol Rep. 2014;32:957-964. doi:10.3892/or.2014.3318

42.

Wang

Zhou

Wang

, et al. A prognostic gene signature based on focal adhesion related genes for gliomas and identification of the role of RAP1B in glioma progression. Research Square Platform LLC [Preprint Posted Online October 18, 2023]. doi:10.21203/rs.3.rs-3437819/v1

43.

Liu

Yang

Xiong

Improved prognostic prediction of glioblastoma using a PAS detected from single-cell RNA-seq. J Cancer. 2020;11:3751-3761. doi:10.7150/jca.44034

44.

Liu

Zhang

, et al. Cysteine and glycine-rich protein 2 is crucial for maintaining the malignant phenotypes of gliomas through its action on Notch signalling cascade. Toxicol Appl Pharmacol. 2024;487:116969. doi:10.1016/j.taap.2024.116969

45.

Ookawa

Wanibuchi

Kataoka-Sasaki

, et al. Digital polymerase chain reaction quantification of SERPINA1 Predicts prognosis in high-grade glioma. World Neurosurg. 2018;111:e783-e789. doi:10.1016/j.wneu.2017.12.166

46.

Lei

Wang

Liu

Thrombospondin-1 is a prognostic biomarker and is correlated with tumor immune microenvironment in glioblastoma. Oncol Lett. 2020;21:22. doi:10.3892/ol.2020.12283

47.

Daubon

Léon

Clarke

, et al. Deciphering the complex role of thrombospondin-1 in glioblastoma development. Nat Commun. 2019;10:1146. doi:10.1038/s41467-019-08480-y

48.

Ventero

Fuentes-Baile

Quereda

, et al. Radiotherapy resistance acquisition in Glioblastoma. Role of SOCS1 and SOCS3. PLoS ONE. 2019;14:e0212581. doi:10.1371/journal.pone.0212581

49.

Chen

, et al. SOCS1: a potential diagnostic and prognostic marker for aggressive gliomas and a new target for immunotherapy. Medicine (Baltimore). 2024;103:e40632. doi:10.1097/md.0000000000040632

50.

Lucca

Lerner

Park

, et al. Differential expression of the T-cell inhibitor TIGIT in glioblastoma and MS. Neurol Neuroimmunol Neuroinflammation. 2020;7:1-20. doi:10.1212/nxi.0000000000000712

51.

Choi

Lee

Park

, et al. Human allogenic γδ T cells kill patient-derived glioblastoma cells expressing high levels of DNAM-1 ligands. Oncoimmunology. 2022;11:2138152. doi:10.1080/2162402x.2022.2138152

52.

Sun

, et al. Interrogating glioma-M2 macrophage interactions identifies Gal-9/Tim-3 as a viable target against PTEN -null glioblastoma. Sci Adv. 2022;8:eabl5165. doi:10.1126/sciadv.abl5165

53.

Karagkouni

Paraskevopoulou

Tastsoglou

, et al. DIANA-LncBase v3: indexing experimentally supported miRNA targets on non-coding transcripts. Nucleic Acids Res. 2019;48:D101-D110. doi:10.1093/nar/gkz1036

54.

Liu

Zhou

Yang

JH.

starBase v2.0: decoding miRNA-ceRNA, miRNA-ncRNA and protein–RNA interaction networks from large-scale CLIP-Seq data. Nucleic Acids Res. 2014;42:D92-97. doi:10.1093/nar/gkt1248

55.

Cheng

Yan

Zhang

, et al. Single-cell transcriptomics reveals immunosuppressive microenvironment and highlights tumor-promoting macrophage cells in Glioblastoma. PLoS ONE. 2025;20:e0312764. doi:10.1371/journal.pone.0312764

56.

Chen

Gao

, et al. MIR4435-2HG in exosomes promotes gastric carcinogenesis by inducing M2 polarization in macrophages. Front Oncol. 2022;12:1017745. doi:10.3389/fonc.2022.1017745

57.

Chen

Han

, et al. Long Noncoding RNA MIR4435-2HG suppresses colorectal cancer initiation and progression by reprogramming neutrophils. Cancer Immunol Res. 2022;10:1095-1110. doi:10.1158/2326-6066.CIR-21-1011

58.

Zhang

Xue

, et al. Integration of single-cell and bulk RNA sequencing data reveals that CYTOR is a potential prognostic and immunotherapeutic response marker for skin cutaneous melanoma. J Cancer. 2024;15:3890-3902. doi:10.7150/jca.94823

Comprehensive ceRNA Profiling Uncovers Clinically Relevant Hub lncRNAs in Glioblastoma

Abstract

Keywords

Introduction

Methods

Data acquisition and processing

Data normalization and filtering

Differential expression analysis

Integration of experimentally validated interactions

Construction of the GBM-specific ceRNA network

Survival analysis

Expression analysis

Functional enrichment analysis

Statistical analysis and software environment

Results

Data acquisition, normalization and quality control for downstream analysis

Differential expression analysis of genes between different databases

ceRNA triplet (lncRNA -| miRNA -> mRNA) construction

Pathway enrichment analysis using GO and KEGG databases

Expression and survival analysis of ceRNA hub lncRNAs

Shared mRNA regulation between hub lncRNAs (CYTOR and MIR4435-2HG) with a prognostic potential

Co-expression analysis between CYTOR and MIR4435-2HG in TCGA-GBM cohort

Discussion

Footnotes

Acknowledgements

ORCID iD

Ethical Considerations

Author Contributions

Funding

Declaration of Conflicting Interests

Data Availability Statement

References