Abstract
BACKGROUND: Developments in high-throughput genomic technologies have led to improved understanding of the molecular underpinnings of esophageal squamous cell carcinoma (ESCC). However, there is currently no model that combines the clinical features and gene expression signatures to predict outcomes. METHODS: We obtained data from the GSE53625 database of Chinese ESCC patients who had undergone surgical treatment. The R packages, Limma and WGCNA, were used to identify and construct a co-expression network of differentially expressed genes, respectively. The Cox regression model was used, and a nomogram prediction model was constructed. RESULTS: A total of 3654 differentially expressed genes were identified. Bioinformatics enrichment analysis was conducted. Multivariate analysis of the clinical cohort revealed that age and adjuvant therapy were independent factors for survival, and these were entered into the clinical nomogram. After integrating the gene expression profiles, we identified a "2-gene score" associated with overall survival. The combinational model is composed of clinical data and gene expression profiles. The C-index of the combined nomogram for predicting survival was statistically higher than the clinical nomogram. The calibration curve revealed that the combined nomogram and actual observation showed better prediction accuracy than the clinical nomogram alone. CONCLUSIONS: The integration of gene expression signatures and clinical variables produced a predictive model for ESCC that performed better than those based exclusively on clinical variables. This approach may provide a novel prediction model for ESCC patients after surgery.