Humans ; Artificial Intelligence ; Lung Neoplasms/pathology* ; Mediastinum/pathology* ; Lymph Nodes/pathology* ; Neoplasm Staging ; Lymph Node Excision

Objective    To establish the gene-based esophageal cancer (ESCA) risk score prediction models via whole transcriptome analysis to provide ideas and basis for improving ESCA treatment strategies and patient prognosis. Methods    RNA sequencing data of esophageal squamous cell carcinoma (ESCC), esophageal adenocarcinoma (EAC) and adjacent tissues were obtained from The Cancer Genome Atlas database. The edgeR method was used to screen out the differential genes between ESCA tissue and normal tissue, and the key genes affecting the survival status of ESCC and EAC patients were initially identified through univariate Cox regression analysis. The least absolute shrinkage and selection operator regression analysis and multivariate Cox regression analysis were used to further screen genes and establish ESCC and EAC risk score prediction models. Results    The risk score prediction models were the independent prognostic factors for ESCA, and the risk score was significantly related to the survival status of patients. In ESCC, the risk score was related to T stage. In EAC, the risk score was related to lymph node metastasis, distant metastasis and clinical stage. The constructed nomogram based on risk score showed good predictive ability. In ESCC, the risk score was related to tumor immune cell infiltration and the expression of immune checkpoint genes. However, this feature was not obvious in EAC. Conclusion 聽 聽The ESCC and EAC risk score prediction models have shown good predictive capabilities, which provide certain inspiration and basis for optimizing the management of ESCA and improving the prognosis of patients.