An In Silico Analysis Reveals an EMT-Associated Gene Signature for Predicting Recurrence of Early-Stage Lung Adenocarcinoma

Abstract

Background: The potential micrometastasis tends to cause recurrence of lung adenocarcinoma (LUAD) after surgical resection and consequently leads to an increase in the mortality risk. Compelling evidence has suggested the underlying mechanisms of tumor metastasis could involve the activation of an epithelial-mesenchymal transition (EMT) program. Hence, the objective of this study was to develop an EMT-associated gene signature for predicting the recurrence of early-stage LUAD. Methods: The mRNA expression data of patients with early-stage LUAD were downloaded from Gene Expression Omnibus (GEO) and The Cancer Genome Atlas (TCGA) available databases. Gene Set Variation Analysis (GSVA) was first performed to provide an assessment of EMT phenotype, whereas Weighted Gene Co-expression Network Analysis (WGCNA) was constructed to determine EMT-associated key modules and genes. Based on the genes, a novel EMT-associated signature for predicting the recurrence of early-stage LUAD was identified using a least absolute shrinkage and selection operator (LASSO) algorithm and a stepwise Cox proportional hazards regression model. Kaplan-Meier survival analysis, receiver operating characteristic (ROC) curves and Cox regression analyses were used to estimate the performance of the identified gene signature. Results: GSVA revealed diverse EMT states in the early-stage LUAD. Further correlation analyses showed that the EMT states presented high correlations with several hallmarks of cancers, tumor purity, tumor microenvironment cells, and immune checkpoint genes. More importantly, Kaplan-Meier survival analyses indicated that patients with high EMT scores had worse recurrence-free survival (RFS) and overall survival (OS) than those with low EMT scores. A novel 5-gene signature (AGL, ECM1, ENPP1, SNX7, and TSPAN12) was established based on the EMT-associated genes from WGCNA and this signature successfully predicted that the high-risk patients had a higher recurrence rate compared with the low-risk patients. In further analyses, the signature represented robust prognostic values in 2 independent validation cohorts (GEO and TCGA datasets) and a combined GEO cohort as evaluated by Kaplan-Meier survival (P-value