DNA damage repair gene signature model for predicting prognosis and chemotherapy outcomes in lung squamous cell carcinoma

Xinshu Wang,Zhiyuan Huang,Lei Li,Guangxue Wang,Lin Dong,Qinchuan Li,Jian Yuan,Yunhui Li
DOI: https://doi.org/10.1186/s12885-022-09954-x
IF: 4.638
2022-01-01
BMC Cancer
Abstract:Background Lung squamous cell carcinoma (LUSC) is prone to metastasis and likely to develop resistance to chemotherapeutic drugs. DNA repair has been reported to be involved in the progression and chemoresistance of LUSC. However, the relationship between LUSC patient prognosis and DNA damage repair genes is still unclear. Methods The clinical information of LUSC patients and tumour gene expression level data were downloaded from the TCGA database. Unsupervised clustering and Cox regression were performed to obtain molecular subtypes and prognosis-related significant genes based on a list including 150 DNA damage repair genes downloaded from the GSEA database. The coefficients determined by the multivariate Cox regression analysis and the expression level of prognosis-related DNA damage repair genes were employed to calculate the risk score, which divided LUSC patients into two groups: the high-risk group and the low-risk group. Immune viability, overall survival, and anticarcinogen sensitivity analyses of the two groups of LUSC patients were performed by Kaplan–Meier analysis with the log rank test, ssGSEA and the pRRophetic package in R software. A time-dependent ROC curve was applied to compare the survival prediction ability of the risk score, which was used to construct a survival prediction model by multivariate Cox regression. The prediction model was used to build a nomogram, the discriminative ability of which was confirmed by C-index assessment, and its calibration was validated by calibration curve analysis. Differentially expressed DNA damage repair genes in LUSC patient tissues were retrieved by the Wilcoxon test and validated by qRT–PCR and IHC. Result LUSC patients were separated into two clusters based on molecular subtypes, of which Cluster 2 was associated with worse overall survival. A prognostic prediction model for LUSC patients was constructed and validated, and a risk score calculated based on the expression levels of ten DNA damage repair genes was employed. The clinical utility was evaluated by drug sensitivity and immune filtration analyses. Thirteen-one genes were upregulated in LUSC patient samples, and we selected the top four genes that were validated by RT–PCR and IHC. Conclusion We established a novel prognostic model based on DNA damage repair gene expression that can be used to predict therapeutic efficacy in LUSC patients.
What problem does this paper attempt to address?