Abstract
BACKGROUND: Glioblastoma (GBM) is the most common and aggressive malignant neoplasm in the central nervous system. Apoptosis is crucial in the genesis, progression, and management of tumors. Nevertheless, the influence of apoptosis-associated genes on GBM prognosis is unclear. METHODS: Transcriptome data and single-cell sequencing data were obtained from TCGA, CGGA, and GEO databases. Differential genes related to apoptosis were screened using the limma software, and an apoptosis-related gene prognostic model (apoptosis signature [AS] model) was constructed through univariate Cox analysis under the optimization of 101 machine learning algorithm combinations. Validation analyses were conducted using bioinformatics tools. RESULTS: A notable divergence in the expression levels of genes associated with programed cell death was identified when comparing GBM neoplastic tissues to their surrounding non-neoplastic counterparts. They were closely related to the prognosis of GBM patients. BRCA1, CHEK2, and IKBKE genes exhibited elevated levels of expression within neoplastic tissues and were identified as risk factors for prognosis, while ZMYND11, MAPK8, and RPS3 genes were highly expressed in adjacent nontumor tissues as protective factors. The AS model demonstrated good predictive performance across multiple datasets, showing a higher concordance index (C-index) value compared to conventional indicators of outcome. Moreover, the correlation coefficient between HSPB1 and the risk score associated with the AS model was positive, with a value of 0.75 (p < 2.2e-16). CONCLUSIONS: An apoptosis-related gene prognostic model (AS model) with high predictive performance was constructed and had close associations with the tumor immune microenvironment and intercellular communication. The HSPB1 had a good predictive effect on GBM prognosis.