Identification of a Nomogram with an Autophagy-Related Risk Signature for Survival Prediction in Patients with Glioma

Abstract

Background: Glioma is a common type of tumor in the central nervous system characterized by high morbidity and mortality. Autophagy plays vital roles in the development and progression of glioma, and is involved in both normal physiological and various pathophysiological progresses.

Patients and methods: A total of 531 autophagy-related genes (ARGs) were obtained and 1738 glioma patients were collected from three public databases. We performed least absolute shrinkage and selection operator regression to identify the optimal prognosis-related genes and constructed an autophagy-related risk signature. The performance of the signature was validated by receiver operating characteristic analysis, survival analysis, clinic correlation analysis, and Cox regression. A nomogram model was established by using multivariate Cox regression analysis. Schoenfeld's global and individual test were used to estimate time-varying covariance for the assumption of the Cox proportional hazard regression analysis. The R programming language was used as the main data analysis and visualizing tool.

Results: An overall survival-related risk signature consisting of 15 ARGs was constructed and significantly stratified glioma patients into high- and low-risk groups (P < 0.0001). The area under the ROC curve of 1-, 3-, 5-year survival was 0.890, 0.923, and 0.889, respectively. Univariate and multivariate Cox analyses indicated that the risk signature was a satisfactory independent prognostic factor. Moreover, a nomogram model integrating risk signature with clinical information for predicting survival rates of patients with glioma was constructed (C-index=0.861±0.024).

Conclusion: This study constructed a novel and reliable ARG-related risk signature, which was verified as a satisfactory prognostic marker. The nomogram model could provide a reference for individually predicting the prognosis for each patient with glioma and promoting the selection of optimal treatment.

Keywords: autophagy; glioma; nomogram; prognosis; risk signature.

Figure 1

Study flowchart.

Figure 2

Establishment of the risk signature with fifteen ARGs in the TCGA dataset. (A). Log (Lambda) value of the 66 ARGs in LASSO regression model. (B). The most proper log (Lambda) value in LASSO regression model. (C). The coefficients of selected genes. (D). The risk scores trend and patients’ survival distribution in TCGA dataset. (E). Receiver operating characteristic curves (ROC) for predicting 1-, 3-, 5-year survival of glioma patients.(F–I). The prognostic value of the ARG-related risk signature showed by Kaplan-Meier survival curves between low-risk group (n = 281) and high-risk group (n = 281) both in all grades and each grade.

Figure 3

Biological processes and KEGG pathway enrichment analyses performed by GSVA in TCGA dataset. (A). GSVA for biological processes analysis; (B). GSVA for KEGG pathways analysis. (|log2 (fold change) | >0.2 and adjusted P < 0.05).

Figure 4

Associations between the ARG-related risk signature and clinicopathologic features in glioma patients in TCGA dataset. (A). Associations between the increasing risk score and clinicopathologic information (age, gender, WHO grade, histological classification, IDH mutation, and 1p/19q codeletion status) of 562 patients. (B–E). The expression levels of approved 15 ARGs in different cohorts stratified by age, WHO grade, IDH mutation, and 1p/19q codeletion status subtypes. ***P < 0.001, ****P < 0.0001.

Figure 5

Distribution of the risk scores in different clinical groups and analyses of independent prognostic significance of the risk signature.(A–F). Distribution of the risk scores for different cohorts stratified by age, gender, WHO grade, histological classification, IDH mutation, and 1p/19q codeletion status subgroups showed by box plots. A, astrocytoma; OA, oligoastrocytoma; OD, oligodendroglioma; GBM, glioblastoma multiforme.(G–H). Results of univariate and multivariate Cox regression analyses in the TCGA training set.

Figure 6

Validation of the expression levels of ARGs. (A). The relative mRNA expression levels of identified 15 ARGs between human brain and glioma samples.(B). Representative Western blots of NAMPT and CTSB protein expression levels in normal human astrocytes (NHA) and human glioma cell lines.(C, D). Relative quantification of Western blots shown in (g). One-way ANOVA for multi-group comparisons: NS, non-significant, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Figure 7

Construction and validation of a prognostic nomogram model. (A). A nomogram for predicting the 1-, 3-, and 5- year survival rates of patients with glioma. (B–D). Schoenfeld’s individual and global test for estimating time-varying covariance. (B) Age; (C) Grade; (D) Risk score.(E). ROC curves were used for evaluating the efficiency of the nomogram in the TCGA dataset.(F–H). The calibration plots for predicting patients’ survival at 1- (F), 3- (G), or 5- (H) year in the TCGA dataset.