Association of M2 macrophages with EMT in glioma identified through combination of multi-omics and machine learning

• Published in: Heliyon Vol. 10; no. 15; p. e34119
• Main Authors: Feng, Peng, Liu, Shangyu, Yuan, Guoqiang, Pan, Yawen
• Format: Journal Article
• Language: English
• Published: England Elsevier Ltd 15.08.2024; Elsevier
• Subjects: ATMs; Glioma; Machine learning; Prognosis; Single cells; Prognosis; Glioma; Single cells; Machine learning; ATMs
• ISSN: 2405-8440; 2405-8440
• DOI: 10.1016/j.heliyon.2024.e34119

The incidence of glioma, a prevalent brain malignancy, is increasing, particularly among the elderly population. This study aimed to elucidate the clinical importance of epithelial–mesenchymal transition (EMT) in gliomas and its association with malignancy and prognosis. The incidence of glioma, particularly among elderly individuals, is on the rise. The malignancy of glioma is determined not only by the oncogenic properties of tumor cells but also by the composition of the tumor microenvironment, which includes immune system macrophages. The prevalence of M2-type macrophages typically fosters tumor progression, yet the underlying mechanism remains elusive. Our study explored the clinical importance of epithelial–mesenchymal transition (EMT) in gliomas and its association with malignancy and prognosis. Our study used the gene set variation analysis (GSVA) algorithm to classify different levels of EMT activation based on the transcriptomic and multi-omics data. Machine learning (ML) and single-cell analysis were integrated into our model for comprehensive analysis. A predictive model was constructed and in vitro experiments were performed to validate our findings. Our study classified 1,641 samples into two clusters based on EMT activation: the EMT-hot group and the EMT-cold group. The EMT-hot group had elevated copy number loss, tumor mutational burden (TMB), and a poorer survival rate. Conversely, the EMT-cold group showed a better survival rate, likely attributed to lower stromal and immune cell scores, as well as decreased expression of human leukocyte antigen-related genes. Driving genes were identified through weighted gene coexpression network analysis (WGCNA) and dimensionality reduction techniques. These genes were then utilized in the construction of a prognostic model using ML and protein-protein interaction (PPI) network analysis. Furthermore, the impact of the core genes identified through single-cell analysis on glioma prognosis was examined. Our research underscores the efficacy of our model in predicting glioma prognosis and elucidates the connection between the M2 macrophages and EMT. Additionally, core genes such as LY96, C1QB, LGALS1, CSPG5, S100A8, and CHGB were identified as pivotal for mediating the occurrence of EMT induced by M2 macrophages.