Establishment and validation of a prognostic prediction model for glioma based on key genes and clinical factors.

Abstract

Background: Glioma is a common brain tumour that is associated with poor prognosis. Immunotherapy has shown significant potential in the treatment of gliomas. Herein, we proposed a new prognostic risk model based on immune- and mitochondrial energy metabolism-related differentially expressed genes (IR&MEMRDEGs) to enhance the accuracy of prognostic assessment in patients with glioma.

Methods: Data from samples from 671 glioma patients and 5 normal controls with available follow-up data and prognostic outcomes were downloaded from the Gene Expression Omnibus (GEO) and The Cancer Genome Atlas (TCGA) databases. All data were downloaded on 13 November 2023. IR&MEMRDEGs were screened from the GeneCards website and published literature. Prognostic prediction models were constructed and analysed using Cox and Least Absolute Shrinkage and Selection Operator (LASSO) regression, Kaplan-Meier (KM) curve, and receiver operating characteristic (ROC) curve analyses. Single-sample gene set enrichment analysis (ssGSEA) was further performed to ascertain the percentage of immune cell infiltration in the glioma specimens.

Results: Bioinformatics analysis of the GEO and TCGA databases identified eleven MEMRDEGs with dysregulated expression in gliomas: EIF4EBP1, TP53, IDH1, PRKCZ, CD200, GPI, PGM2, PKLR, AK2, ATP4A, and ALDH3B1. Further analysis identified EIF4EBP1, TP53, IDH1, PRKCZ, CD200, GPI, PGM2, AK2, and ALDH3B1 as separate predictive factors for glioma, among which PGM2 and AK2 exhibited superior accuracy [area under the ROC curve (AUC) >0.9], while EIF4EBP1, TP53, IDH1, PRKCZ, GPI, and ALDH3B1 demonstrated slightly lower accuracy (0.7< AUC <0.9), and CD200 displayed poor accuracy (0.5< AUC <0.7). Among these genes, the levels of AK2, ALDH3B1, EIF4EBP1, GPI, IDH1, PGM2, and TP53 were significantly higher in the high-risk group (HRG) compared with the low-risk group (LRG) (P<0.001), indicating a negative association with patient prognosis. In contrast, CD200 and PRKCZ were significantly downregulated in the HRG compared to the LRG (P<0.05), indicating a potential correlation with patient outcomes. Subsequently, prognostic models were constructed based on IR&MEMRDEG and MEMRDEGs to anticipate the outcomes of glioma patients, while the predictive efficacy of the model was validated via KM and ROC curve analysis. The results revealed that EIF4EBP1, TP53, IDH1, PRKCZ, GPI, PGM2, ALDH3B1, and AK2 had superior accuracy in predicting glioma prognosis. The ssGSEA results showed that only IDH1 was negatively linked to the amount of immune cell infiltration in the LRG, while displaying a positive connection in the HRG (r value>0), indicating that the expression levels of IDH1 may have a distinct influence on the tumour immune microenvironment.

Conclusions: The present study confirmed the significant predictive value of IDH1 for glioma prognosis, which may guide immunotherapy for glioma treatment.