Construction of a prognostic model for advanced non-small cell lung cancer using combined analysis of public databases

Non-small cell lung cancer (NSCLC) has become a global health problem owing to its high incidence and poor treatment efficacy. This study aimed to establish a prognostic model for advanced NSCLC using The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) to evaluate the prognosis. Based on TCGA database, we obtained a dataset of 191 advanced NSCLC and 108 normal tissue samples after screening. We screened the differential expression between advanced NSCLC and normal tissues using the "limma" software package in R and identified differential expression genes. Weighted gene co-expression network analysis of differential genes was performed and the Gene Ontology database and Kyoto Encyclopedia of Genes and Genomes were used for pathway enrichment analysis of the key genes in the core modules. Finally, univariate Cox regression analysis was used to determine genes associated with prognosis, and multivariate Cox regression analysis was used to construct a prognostic model. We used the receiver operating characteristic curve (ROC), Kaplan-Meier curve, and multivariate Cox regression analysis to test the prognostic value of the model. The GEO database (GSE68465) was used to screen a cohort of clinical and gene expression data from 69 patients to further validate the prognostic value of the model. We constructed a five-gene ( CKS1B , SLC7A5 , SPATS2 , SPDL1 , and CHEK2 ) prognostic model to predict overall survival in patients with advanced NSCLC. Kaplan-Meier analysis, area under the ROC curve (AUC), and C-index had good predictive power for gene signatures in TCGA dataset (P<0.0001; C-index 0.710; 95% confidence interval (0.681–0.738); AUC at 1 year, 2 years, and 3 years were 0.76, 0.75, and 0.73, respectively). The risk score was an independent risk factor for survival in patients with advanced NSCLC, after adjusting for other routine clinical parameters. Finally, we constructed a nomogram that integrated gene signatures and clinicopathological parameters. The risk score model constructed from the five survival-related genes can be used as prognostic biomarkers. The nomogram was closely related to the overall survival of the patients. This result may help future individualized treatment and clinical decision making.