Top-down stepwise refinement identifies coding and noncoding RNA-associated epigenetic regulatory maps in malignant glioma.

With the emergence of the molecular era and retreat of the histology epoch in malignant glioma, it is becoming increasingly necessary to research diagnostic/prognostic/therapeutic biomarkers and their related regulatory mechanisms. While accumulating studies have investigated coding gene-associated biomarkers in malignant glioma, research on comprehensive coding and noncoding RNA-associated biomarkers is lacking. Furthermore, few studies have illustrated the cross-talk signalling pathways among these biomarkers and mechanisms in detail. Here, we identified DEGs and ceRNA networks in malignant glioma and then constructed Cox/Lasso regression models to further identify the most valuable genes through stepwise refinement. Top-down comprehensive integrated analysis, including functional enrichment, SNV, immune infiltration, transcription factor binding site, and molecular docking analyses, further revealed the regulatory maps among these genes. The results revealed a novel and accurate model (AUC of 0.91 and C-index of 0.84 in the whole malignant gliomas, AUC of 0.90 and C-index of 0.86 in LGG, and AUC of 0.75 and C-index of 0.69 in GBM) that includes twelve ncRNAs, 1 miRNA and 6 coding genes. Stepwise logical reasoning based on top-down comprehensive integrated analysis and references revealed cross-talk signalling pathways among these genes that were correlated with the circadian rhythm, tumour immune microenvironment and cellular senescence pathways. In conclusion, our work reveals a novel model where the newly identified biomarkers may contribute to a precise diagnosis/prognosis and subclassification of malignant glioma, and the identified cross-talk signalling pathways would help to illustrate the noncoding RNA-associated epigenetic regulatory mechanisms of glioma tumorigenesis and aid in targeted therapy.