Work

Machine learning-based integration develops a hypoxia-derived signature for improving outcomes in glioma

Year: 2025

Type: article

Abstract: Background: The growth of glioma is frequently accompanied by a hypoxic microenvironment, which contributes to its progression and unfavorable prognosis. Nevertheless, the clinical implications of hypoxia have not been extensively investigated. Methods: A comprehensive collection of over 3000 glioma samples was acquired from publicly available databases. The identification of a hypoxia-cluster was accomplished through the utilization of Consensus Cluster Plus across six cohorts. Subsequently, a hypoxia-derived signature was formulated by employing 100 machine learning algorithms on the expression profiles of 11 genes within six glioma cohorts. To further investigate the expression of these 11 genes and certain hub molecules, RT-qPCR and IHC were conducted in an in-house cohort. Results：Two hypoxia-associated subtypes (C1 and C2) were identified in glioma, exhibiting notable prognostic and molecular disparities. Subtype C2 gliomas demonstrated an abundance of hypoxia-related pathways, increased presence of immune and stromal cells, elevated expression of immune checkpoint genes and receptors, and a more unfavorable prognosis compared to those in C1. Employing a machine learning-based integrative approach, we established an 11-gene signature capable of predicting distant clinical outcomes across six cohorts. This signature was subsequently validated through RT-qPCR analysis in an in-house cohort, and effectively categorizes patients into high-risk and low-risk groups, demonstrating consistent and robust performance in predicting both overall and relapse-free survival. Moreover, the risk score exhibits significantly higher accuracy compared to conventional clinical variables, molecular characteristics, and 100 previously published signatures. Glioma in the high-risk group exhibits elevated levels of CD163, PD1, HIF1A, and PD-L1 expression. Conclusion: Ultimately, we developed a novel hypoxia-related classification that provides the potential to guide treatment decisions and a reliable prognostic tool.

Source: iScience

DOI: 10.1016/j.isci.2025.112023