Four differentially expressed genes can predict prognosis and microenvironment immune infiltration in lung cancer: a study based on data from the GEO

Abstract

Abstract Background Lung cancer is among the major diseases threatening human health. Although the immune response plays an important role in tumor development, its exact mechanisms are unclear. Materials and methods Here, we used CIBERSORT and ESTIMATE algorithms to determine the proportion of tumor-infiltrating immune cells (TICs) as well as the number of immune and mesenchymal components from the data of 474 lung cancer patients from the Gene Expression Omnibus database. And we used data from The Cancer Genome Atlas database (TCGA) for validation. Results We observed that immune, stromal, and assessment scores were only somewhat related to survival with no statistically significant differences. Further investigations revealed these scores to be associated with different pathology types. GO and KEGG analyses of differentially expressed genes revealed that they were strongly associated with immunity in lung cancer. In order to determine whether the signaling pathways identified by GO and KEGG signaling pathway enrichment analyses were up- or down-regulated, we performed a gene set enrichment analysis using the entire matrix of differentially expressed genes. We found that signaling pathways involved in hallmark allograft rejection, hallmark apical junction, hallmark interferon gamma response, the hallmark P53 pathway, and the hallmark TNF-α signaling via NF-ĸB were up-regulated in the high-ESTIMATE-score group. CIBERSORT analysis for the proportion of TICs revealed that different immune cells were positively correlated with the ESTIMATE score. Cox regression analysis of the differentially expressed genes revealed that CPA3, C15orf48, FCGR1B, and GNG4 were associated with patient prognosis. A prognostic model was constructed wherein patients with high-risk scores had a worse prognosis (p < 0.001 using the log-rank test). The Area Under Curve (AUC)value for the risk model in predicting the survival was 0.666. The validation set C index was 0.631 (95% CI: 0.580–0.652). The AUC for the risk formula in the validation set was 0.560 that confirmed predictivity of the signature. Conclusion We found that immune-related gene expression models could predict patient prognosis. Moreover, high- and low-ESTIMATE-score groups had different types of immune cell infiltration.