A Novel Prognosis Signature Based on Ferroptosis-Related Gene DNA Methylation Data for Lung Squamous Cell Carcinoma

Xuequan Wang; Yinnan Meng; Changxin Liu; Haihua Yang; Suna Zhou

doi:10.1155/2022/9103259

A Novel Prognosis Signature Based on Ferroptosis-Related Gene DNA Methylation Data for Lung Squamous Cell Carcinoma

J Oncol. 2022 Sep 12:2022:9103259. doi: 10.1155/2022/9103259. eCollection 2022.

Authors

Xuequan Wang¹, Yinnan Meng¹, Changxin Liu², Haihua Yang¹, Suna Zhou^{1

3}

Affiliations

¹ Key Laboratory of Radiation Oncology of Taizhou, Radiation Oncology Institute of Enze Medical Health Academy, Department of Radiation Oncology, Taizhou Hospital Affiliated to Wenzhou Medical University, Taizhou, Zhejiang Province 317000, China.
² Department of Epidemiology and Biostatistics, School of Public Health, Xi'an Jiaotong University Health Science Center, Xi'an, Shaanxi 710061, China.
³ Department of Radiation Oncology, Xi'an No. 3 Hospital, The Affiliated Hospital of Northwest University, Xi'an, Shaanxi 710018, China.

Abstract

Ferroptosis-related genes regulating an iron- and lipid reactive oxygen species (ROS)-dependent form of programmed cell death suggest critical roles for ferroptosis in cancers. However, the prognostic value of ferroptosis-related epigenetic features such as DNA methylation in lung squamous cell carcinoma (LUSC) needs to be studied. Ferroptosis-related genes are collected from the FerrDb database, and the methylation data of these related genes in LUSC methylation data downloaded from the TCGA are retrieved. The DNA methylation data (362 LUSC samples) were analyzed to screen prognostic ferroptosis-related methylation sites. After patients with complete overall survival (OS) information were randomly separated into training cohort (n = 200) and validation cohort (n = 162), the least absolute shrinkage and selection operator (LASSO) and the Cox regression were used to establish and validate the prognostic signature. The time-dependent receiver operating characteristic (ROC) and Kaplan-Meier survival curve analyses, Harrell's concordance index (C-index), calibration analysis, and decision curve analysis (DCA) were performed to evaluate the risk signature and related nomogram. A series of other bioinformatics approaches such as mexpress, cbioportal, maftools, string, metascape, TIMER, and Kaplan-Meier survival curve analysis were also used to determine the methylation, mutation status, protein interaction network or functional enrichment, effects on immune cell infiltration, or expression level prognosis of those signature-related genes. A total of 137 DNA methylation sites were identified as prognostic predictors corresponding to 109 ferroptosis-related genes (FRGs). The methylation signature containing 31 methylation sites proved to be superior predictive efficiency in predicting the 1-, 3-, 5-, and 10-year OS. 8 out of 28 signature-related genes were significantly related to OS time or OS state in patients with LUSC. In addition, DUSP1, ZFN36, and ALOX5 methylation status also correlated with pathological M and ALOX5 methylation correlated with pathological N. The prognostic prediction efficiency of T, N, M, and the stage was inferior to that of the DNA methylation signature. LUSC patients in the high-risk group own a significantly larger number of variants of FRGs than those in the low-risk group. In addition, negative or positive correlation patterns were presented among the different infiltrating immune cells with risk scores or signature-related genes in patients with LUSC. The expression level of 15 signature-related genes showed a significant relationship with OS of LUSC patients. A novel prognostic nomogram survival model containing 4 factors including age, pathologic T, stage, and risk group was constructed and validated, AndC-index, decision curve analysis (DCA), and calibration analysis demonstrated its excellent predictive performance. The FRG DNA methylation data-based prognostic model acts as a powerful prognostic prediction indicator in LUSC patients and is advantageous over the traditional model based on T, N, M, and stage.