Establishment of a cell senescence related prognostic model for predicting prognosis in glioblastoma

Background: Glioblastoma (GBM) is highly malignant and has a worse prognosis with age, and next-generation sequencing (NGS) provides us with a huge amount of information about GBM. Materials and Methods: Through the enrichment scores of cell senescence-related pathways, we constructed a consensus matrix and mined molecular subtypes and explored the differences in pathological, immune/pathway and prognostic. Also we identified key genes related to cell senescence characteristics using least absolute shrinkage and selection operator (Lasso) regression and univariate COX regression analysis models. The use of risk factor formats to construct clinical prognostic models also explored the differences in immunotherapy/chemotherapy within the senescence-related signatures score (SRS.score) subgroups. Decision trees built with machine learning to identify the main factors affecting prognosis have further improved the prognosis model and survival prediction. Results: We obtained seven prognostic-related pathways related to cell senescence. We constructed four different molecular subtypes and found patients with subtype C1 had the worst prognosis. C4 had the highest proportion of patients with IDH mutations. 1005 differentially expressed genes (DEGs) were analyzed, and finally 194 Risk genes and 38 Protective genes were obtained. Eight key genes responsible for cell senescence were finally identified. The clinical prognosis model was established based on SRS.score, and the prognosis of patients with high SRS.score was worse. SRS.score and age were the vital risk factors for GBM patients through decision tree model mining. Conclusion: We constructed a clinical prognosis model that could provide high prediction accuracy and survival prediction ability for adjuvant treatment of patients with GBM.