Weighted Gene Co-Expression Network Analysis Identifies an Immunogenic Cell Death Signature to Predict Therapeutic Responses and Prognosis of Glioblastoma

Abstract

Background: Induction of immunogenic cell death (ICD) breaks down the immunosuppressive tumor microenvironment (TME) and controls tumor progression, but the correlation between glioblastoma (GBM) and ICD is unclear. Therefore, this study aims to investigate the potential prognostic value of ICD-associated genes in GBM. Methods: We collected 34 ICD-related genes from various sources. Utilizing public databases, we extracted relevant GBM data and delineated prognosis-related ICD gene modules using weighted gene co-expression network analysis (WGCNA). Least absolute shrinkage and selection operator (LASSO) algorithm was employed to develop a risk model, whose accuracy was confirmed by including an independent Gene Expression Omnibus (GEO) dataset. The biological functions and pathways associated with these signals were analyzed by performing enrichment analysis, and the tumor immune infiltration capacity was evaluated. The R package oncoPredict was used to infer the drug sensitivity of patients in different risk groups using data from the Genomics of Drug Sensitivity in Cancer 2 (GDSC2) database with expression profiling. Results: Thirty-four ICD-associated genes were differentially expressed in GBM samples and two gene modules significantly associated with prognosis were identified. Based on these gene modules, vitamin D receptor (VDR) and cell death-inducing DFF45-like effector B (CIDEB) were identified as two signature genes for the prognostic prediction of GBM. Subsequently, multivariate Cox analysis confirmed the validity of this signature as an independent factor for evaluating overall survival in GBM. Receiver operating characteristic (ROC) curves also supported an effective prediction of the signature (1-year area under the ROC curve (AUC): 0.667; 3-year AUC: 0.727; 5-year AUC: 0.762). We observed that the high-risk group had higher immune cell infiltration and sensitivity to some drugs. Conclusions: This work developed a novel ICD-related prognostic model for GBM patients. Our findings highlight the potential of using ICD as a promising prognosis indicator in GBM, contributing to the current understanding of the intricate interplay between ICD and tumor microenvironment.