What if glioblastoma prognosis could be read through a molecular tug-of-war between circular RNAs, microRNAs, and genes? This study builds that network and highlights two genes tied to patient survival. Circular RNA plays a vital role in tumor pathogenesis as a competitive endogenous RNA, or ceRNA.

In glioblastoma, the study uses that idea to motivate a network for mechanisms, biomarkers, and therapeutic drugs. The need is especially urgent because glioblastoma has high aggressiveness, a high fatality rate, and poor prognosis. The proposed network is intended to support further study of its mechanism and possible treatment targets.

Glioblastoma is the most frequent primary malignant brain tumor in adults and is often rapidly fatal. Its complications also impose a substantial economic burden and medical strain around the world. Despite advances in current treatments, their impact on glioblastoma remains limited, possibly because its pathogenesis is insufficiently understood.

More studies are therefore needed to investigate mechanisms and identify novel biomarkers for prognosis and treatment outcomes. CircRNAs are more stable than linear RNAs because they are unaffected by exonucleases, and they often show tissue- or developmental-stage-specific expression.

These properties imply biological functions in diseases including glioblastoma. Under the competing endogenous RNA hypothesis, circRNAs act as microRNA sponges: they bind microRNAs and affect corresponding target genes.

The study therefore hypothesizes that microRNA expression is negatively correlated with circRNA and messenger RNA expression. Although ceRNA is important in tumorigenesis, the roles of the circRNA–messenger RNA–microRNA network in glioblastoma remain unclear. The research uses comprehensive analysis of multiple databases.

It examines differentially expressed genes in relation to glioblastoma overall survival, constructs a prognosis-related subnetwork, and proposes drugs that might be potential therapeutic agents. The study downloaded circRNA, microRNA, and messenger RNA expression profiles from three GEO datasets: GSE165926, GSE103229, and GSE90886.

GSE165926 contained circRNA data from four normal brains and twelve glioblastoma tissues. GSE103229 contained microRNA profiles from five normal brain tissues and five glioblastoma tissues. GSE90886 included messenger RNA profiles from nine normal tissues and nine glioblastoma tissues.

TCGA data added five non-tumor brains and one hundred sixty-nine glioblastoma samples for survival analysis and clinical information. The LIMMA package in R identified differentially expressed microRNAs, circRNAs, and messenger RNAs between glioblastomas and non-tumor brain tissues.

The thresholds were absolute log two fold change of at least two point zero, two point zero, and one point zero respectively, with P less than zero point zero five. The differentially expressed genes from GEO and TCGA were divided into upregulated and downregulated groups for further analysis, and Venn diagrams were drawn using the Venn package in R.

The analysis converted differentially expressed circRNA names to mature circRNA names using circBASE, then used CSCD to find microRNA response elements. Only interactions supported by reporter assays, western blots, and quantitative PCR were included.

TargetScan, miRTarBase, and miRDB were used to find microRNA target genes. The analysis retained intersecting messenger RNAs and corresponding microRNAs, then visualized the circRNA–microRNA–messenger RNA network with Cytoscape.

Figure one summarizes the study’s workflow, beginning with circRNA, miRNA, and mRNA microarray datasets processed using the limma package in R. The authors identify differentially expressed RNAs, overlap candidates across datasets, and use interaction databases to predict targeted genes.

The resulting candidates support construction of a ceRNA network, functional enrichment analysis, prognostic-gene identification, ceRNA subnetworks, drug–gene interactions, and meta-analysis validation. The analysis identified two hundred seventy differentially expressed circRNAs: one hundred eighty-three downregulated and eighty-seven upregulated.

It also screened seventy-one differentially expressed microRNAs, including thirty-four downregulated and thirty-seven upregulated. The messenger RNA dataset produced five hundred four differentially expressed messenger RNAs: two hundred forty-two downregulated and two hundred sixty-two upregulated, using an absolute log two fold change threshold of at least one and P less than zero point zero five.

Heatmaps displayed some differentially expressed genes and circRNAs, while structure diagrams showed several differentially expressed circRNAs. Figure three uses two Venn diagrams to show the overlap between datasets.

In panel A, forty-four differential microRNAs are shared between the GSE103229 results and the integrated microRNA set from GSE165926, while twenty-seven and two thousand two hundred twenty-eight remain unique to the respective sets. Panel B shows forty-eight shared target messenger RNAs, supporting the authors’ selection of molecules for the downstream competing endogenous RNA network.

Forty-four of seventy-one differentially expressed microRNAs in GSE103229 matched the integrated microRNAs in GSE165926. Forty-eight of five hundred four differentially expressed messenger RNAs in GSE90886 corresponded with the microRNA-targeted genes. Eleven differentially expressed circRNAs and fifteen differentially expressed messenger RNAs formed seventeen interacting pairs.

Separately, twenty-two differentially expressed circRNAs and eleven differentially expressed microRNAs formed twenty-four interaction pairs. The identified microRNA and messenger RNA overlaps, together with the reported interacting pairs, define the components used in the network analysis.

These components were combined with the miRTarBase databases to construct the circRNA–microRNA–messenger RNA ceRNA network described in the study. Figure four visualizes the competing endogenous RNA, or ceRNA, network linking circRNAs, miRNAs, and mRNAs.

Triangles encode miRNAs, diamonds encode circRNAs, and elliptical nodes encode mRNAs, with light and dark colors indicating downregulation and upregulation respectively. The connecting lines show the reported relationships among these differentially expressed molecules, providing a compact view of how candidate regulatory interactions are organized for further interpretation.

Functional enrichment linked the messenger RNAs to ossification in biological process, the Wnt signalosome in cellular component, and phosphoric diester hydrolase activity in molecular function. KEGG analysis revealed nineteen statistically significant enriched pathways.

The three most enriched were melanogenesis, Cushing syndrome, and the Wnt signaling pathway. Figure six summarizes functional enrichment among mRNAs in the circRNA-associated ceRNA network. Panels A through C organize Gene Ontology results into biological process, cellular component, and molecular function, highlighting terms such as ossification, Wnt signalosome, and phosphoric diester hydrolase activity.

Panel D presents the KEGG enrichment circle plot, including melanogenesis, Cushing syndrome, and the Wnt signaling pathway, helping connect the network’s genes to broader biological pathways. To investigate prognostic ability, fifteen messenger RNAs in the circRNA-associated ceRNA networks were analyzed with the survival package in R.

Two of the fifteen were significantly correlated with overall survival in glioblastoma, using a false discovery rate below zero point zero five. The two messenger RNAs were FZD1 and KLF10, and both were inversely correlated with overall survival.

Their associations with several common immune cell types were also analyzed in the TIMER database. Figure eight links two network mRNAs to both prognosis and the immune environment in glioblastoma. Kaplan–Meier curves in panels A and B show that FZD1 and KLF10 expression were each significantly associated with overall survival, with p-values of zero point zero three five and zero point zero zero three.

Panels C and D then use TIMER scatterplots to examine each gene’s relationship with immune-cell infiltration, extending the survival findings toward possible roles in the GBM immune microenvironment. A meta-analysis initially validated the prognostic ability of FZD1 and KLF10.

It included six FZD1-related articles with nine hundred seventy-two patients and five KLF10-associated articles with six hundred sixty-five patients. High FZD1 expression was significantly associated with poor overall survival, with a hazard ratio of two point five five and P less than zero point zero zero zero zero one.

High KLF10 expression was significantly correlated with inferior overall survival, with a hazard ratio of one point four one and P equal to zero point zero four. The reported heterogeneity was not significant for either analysis, so fixed-effect models were applied.

Funnel-plot analysis found no significant bias. The study used TCGA clinical information to estimate the prognostic values of differentially expressed messenger RNAs and used the survival package in R to find associations with survival status data. A false discovery rate below zero point zero five was statistically significant.

The association of FZD1 and KLF10 with immune infiltration in glioblastoma was analyzed with TIMER, which investigates immune-cell infiltration levels using TCGA data. Cytoscape was used to construct the prognostic ceRNA subnetwork. Figure nine presents the prognostic circRNA–miRNA–mRNA subnetwork as a regulatory map.

Light green diamonds show downregulated circRNAs, dark red triangles show upregulated miRNAs, and light blue ovals identify the downregulated mRNAs KLF10 and FZD1, with connecting lines indicating their reported relationships. The authors also relate FZD1 and KLF10 expression to immune-cell infiltration in glioblastoma, making this network relevant for connecting prognostic RNA regulation with the tumor immune environment.

Differentially expressed messenger RNAs were paired with existing compounds to discover potential therapeutic agents for glioblastoma. DGIdb provides an interface for searching gene lists using drug–gene interaction profiles and potentially druggable genes. The differentially expressed messenger RNAs were uploaded into DGIdb to match them with existing drugs.

The study has important limitations. Tumor samples were not evaluated for glioblastoma phenotypes, which may have influenced expression profiles and prognosis, and survival analyses used online databases with automatic assessment of confounding factors.

Using prognostic value as the primary screening measure may have omitted valuable information. In the meta-analysis, most included studies came from China, and potential publication bias may remain. The ceRNA network, genes, and drugs rely heavily on bioinformatics tools and databases, so their accuracy still needs validation by basic experiments and clinical trials.

The study integrated multiple databases and methods to improve the fidelity of its predictions. The study identifies FZD1 and KLF10 as survival-associated genes within a larger circular RNA network, but the proposed biomarkers and drugs still require validation in experiments and clinical trials.