Research Paper Volume 15, Issue 10 pp 4051—4070

Integrating machine learning and bioinformatics analysis to m6A regulator-mediated methylation modification models for predicting glioblastoma patients’ prognosis and immunotherapy response

Chuanyu Li1, *, , Wangrui Liu1,2, *, , Chengming Liu3, *, , Qisheng Luo1, , Kunxiang Luo1, , Cuicui Wei4, , Xueyu Li1, , Jiancheng Qin1, , Chuanhua Zheng1, , Chuanliu Lan1, , Shiyin Wei1, , Rong Tan1, , Jiaxing Chen1, , Yuanbiao Chen1, , Huadong Huang1, &, , Gaolian Zhang5, , Haineng Huang1, &, , Xiangyu Wang6, ,

  • 1 Department of Neurosurgery, Affiliated Hospital of Youjiang Medical University for Nationalities, Baise 533000, Guangxi, China
  • 2 Department of Interventional Oncology, Renji Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai 200127, China
  • 3 Department of Neurosurgery, The First Affiliated Hospital of Soochow University, Suzhou 215006, Jiangsu, China
  • 4 Department of Outpatient, Affiliated Hospital of Youjiang Medical University for Nationalities, Baise 533000, Guangxi, China
  • 5 Department of Neurosurgery, The First Affiliated Hospital of Guangxi University of Chinese Medicine, Nanning 530023, Guangxi, China
  • 6 Department of Neurosurgery, The First Affiliated Hospital of Jinan University, Guangzhou 510632, Guangdong Province, China
* Equal contribution

Received: August 11, 2022       Accepted: November 30, 2022       Published: May 23, 2023
How to Cite

Copyright: © 2023 Li et al. This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY 3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.


Background: Epigenetic regulations of immune responses are essential for cancer development and growth. As a critical step, comprehensive and rigorous explorations of m6A methylation are important to determine its prognostic significance, tumor microenvironment (TME) infiltration characteristics and underlying relationship with glioblastoma (GBM).

Methods: To evaluate m6A modification patterns in GBM, we conducted unsupervised clustering to determine the expression levels of GBM-related m6A regulatory factors and performed differential analysis to obtain m6A-related genes. Consistent clustering was used to generate m6A regulators cluster A and B. Machine learning algorithms were implemented for identifying TME features and predicting the response of GBM patients receiving immunotherapy.

Results: It is found that the m6A regulatory factor significantly regulates the mutation of GBM and TME. Based on Europe, America, and China data, we established m6Ascore through the m6A model. The model accurately predicted the results of 1206 GBM patients from the discovery cohort. Additionally, a high m6A score was associated with poor prognoses. Significant TME features were found among the different m6A score groups, which demonstrated positive correlations with biological functions (i.e., EMT2) and immune checkpoints.

Conclusions: m6A modification was important to characterize the tumorigenesis and TME infiltration in GBM. The m6Ascore provided GBM patients with valuable and accurate prognosis and prediction of clinical response to various treatment modalities, which could be useful to guide patient treatments.


AUC: area under the curve; CGGA: Chinese Glioma Genome Atlas; CNV: Copy number variation; DEGs: Differential expressed genes; EMT: epithelial-mesenchymal transitions; GGI: Gene-gene interaction; GSVA: Gene set variation analysis; GBM: glioblastoma; OS: overall survival; ROC: Receiver operating characteristic; ssGSEA: single sample Gene Set Enrichment Analysis; TCGA: the Cancer Genome Atlas; TIDE: tumor immune dysfunction and exclusion; TME: tumor microenvironment.