Identification and Validation of Prognostic Risk Model for Female-Specific Lung Adenocarcinoma.

Abstract

Background: Lung adenocarcinoma (LUAD) is a major pathological subtype of non-small cell lung cancer and occurs more commonly in females than other lung cancer subtypes. Studying female-specific oncogenes in LUAD may provide personalized medicine approaches for females with LUAD.

Objective: We aimed to identify the possible female-specific oncogenes of LUAD and understand their potential impact on treatment strategies for specific cancer subgroups.

Methods: The gene expression profiles of LUAD were downloaded from The Cancer Genome Atlas (TCGA) database and the GSE72094 dataset. TCGA database is currently the largest database of cancer genetic information. Female-specific differentially expressed genes (DEGs) were identified by R programming software. Functional annotation of DEGs was conducted based on KEGG pathway enrichment analysis. Univariate and multivariate Cox proportion analyses were applied to construct a prognostic risk score model with the DEGs. Kaplan‒Meier and ROC curves were plotted to validate the predictive effect of the prognostic DEGs signature. Gene set enrichment analysis (GSEA) was applied to identify the potential pathways in the high-risk groups in female LUAD. Finally, the immunohistochemical staining (IHC) was conducted to verify the expression of CABLES1 in human LUAD samples.

Results: We constructed a prognostic signature that includes 12 female-specific DEGs (P < .05). Among them, ABHD6, CABLES1, CXCL5, DNAJB4, EFNB2, HLX, MEOX2, MTMR10, PPFIBP1, and RERG were down-regulated in LUAD, while MFSD6L and SOX9 were up-regulated in LUAD (P < .0001). The Kaplan-Meier, and receiver operator characteristic (ROC) curves revealed efficient and stable prediction of the prognostic signature in the female LUAD patients. It was showed the risk score model has a good predictive effect on the prognosis of female LUAD patients but is not effective for male patients (P < .0001). The ROC curve showed that the areas under the curve (AUC) of first-, third- and fifth-year survival were 0.70, 0.69, and 0.79, respectively, which indicated good sensitivity and specificity of the 12-gene risk score algorithm in predicting the prognosis of female LUAD. GSEA revealed that the high-risk group was significantly enriched in the EMT, E2F targets, Myc targets, G2/M checkpoint, glycolysis, hypoxia, and mTORC1 signaling pathways (P < .05). Immunohistochemical staining showed lower CABLES1 expression was associated with higher pTNM stage in female LUAD but not in male LUAD (P < .05).

Conclusion: Our study constructed and verified a prognostic signature based on 12 female-specific DEGs of LUAD, which could improve the understanding of sex-related risk factors involved in LUAD carcinogenesis and progression, and may provide personalized treatment strategies for female LUAD patients.