Identification and Validation of a Four-Gene Ferroptosis Signature for Predicting Overall Survival of Lung Squamous Cell Carcinoma

Abstract

Background: Lung squamous cell carcinoma (LUSC) represents 30% of all non-small cell lung carcinoma. Targeted therapy is not sufficient for LUSC patients because of the low frequency of targeted-effective mutation in LUSC whereas immunotherapy offers more options for patients with LUSC. We explored a ferroptosis-related prognostic signature that can potentially assess the prognosis and immunotherapy efficacy of LUSC patients. Methods: A total of 502 LUSC patients were downloaded from The Cancer Genome Atlas (TCGA). The external validation data were obtained from the Gene Expression Omnibus (GEO): GSE73403. Then, we identified the candidate genes and constructed the prognostic signature through the Cox survival regression analyses and least absolute shrinkage and selection operator (LASSO). Risk score plot, Kaplan–Meier curve, and ROC curve were used to assess the prognostic power and performance of the model. The CIBERSORT algorithm estimated the fraction of immune cell types. TIDE was utilized to predict the response to immunotherapy. IMvigor210 was used to explore the association between the risk scores and immunotherapy outcomes. A nomogram combined selected clinical characteristics, and the risk scores were constructed. Results: We screened 132 differentially expressed ferroptosis-related genes. According to KEGG and GO pathway analyses, these genes were mainly engaged in the positive regulation of cytokine production, cytokine metabolic process, and oxidoreductase activity. We then constructed a prognostic model via LASSO regression. The proportions of CD8+ T cells, CD4+ activated T cells, and follicular helper T cells were significantly different between low-risk and high-risk groups. TIDE algorithm indicated that low-risk LUSC patients might profit more from immune checkpoint inhibitors. The predictive value of the ferroptosis gene model in immunotherapy response was further confirmed in IMvigor210. Finally, we combined the clinical characteristics with a LASSO regression model to construct a nomogram that could be easily applied in clinical practice. Conclusion: We identified a prognostic model that provides an accurate and objective basis for guiding individualized treatment decisions for LUSC.