Mining multi-omics data for molecular subtyping in gliomas facilitates precise clinical treatment

Abstract

Glioma is the most common malignancy in the skull, accounting for approximately 2% of adult systemic tumors. It is characterized by a high recurrence rate, high disability rate, and poor sensitivity to chemotherapy. Different subtypes of glioma exhibit varying prognosis and chemotherapy sensitivities. Currently, researchers extensively study the molecular classification of glioma based on transcriptomic features, facilitating the evaluation of prognosis. However, identification of tumor subtypes based on the single-layer-omics data have several limitations. In this regard, the present study aimed to identify new subtypes of glioma using three omics datasets for prognosis prediction. As a result, cluster A subtype was identified to be associated with lower activation of cell proliferation and better prognosis, while cluster B subtype exhibited higher infiltration of M2-type macrophages and higher activation of the epithelial-mesenchymal transition (EMT) pathway, potentially leading to a poorer prognosis. Furthermore, we employed the Weighted Gene Co-expression Network Analysis (WGCNA) algorithm and limma R package to identify driver genes for clusters A and B. Consequently, nine genes were identified as gene signatures for cluster A. Based on this finding, we quantified the two subtypes using the single-sample gene set enrichment analysis (ssGSEA) algorithm, where a higher score were linked to elevated tumor mutation burden (TMD) and signaling pathways related to cell proliferation. Low scores indicated enrichment of tumorigenesis-related pathways and poor prognosis. In silico drug screening suggested that AGI-6780 could be an effective compound for high cluster subtypes, based on brain tumor cell lines. Consequently, our study determined that the score can serve as an effective index to predict the prognosis of glioma.