Identification of a prognostic signature derived from immunosuppression-related gene sets in esophageal squamous cell carcinoma via bioinformatics and machine learning

Esophageal squamous cell carcinoma (ESCC) is a digestive tract tumor with a poor prognosis. Immunosuppression is a key driver of tumor immune evasion, but it remains unclear whether immunosuppression-related genes (IRGs) can capture the immunosuppressive microenvironment and predict outcomes in ESCC. ESCC transcriptomic and clinical information were sourced from TCGA and GEO, while immunosuppression-related genes (IRGs) were obtained from the HisgAtlas database. Differentially expressed genes (DEGs) were identified, and their overlap with IRGs defined differentially expressed immunosuppression-related genes (IDEGs). Prognostic models were constructed using univariate Cox regression and three machine learning approaches (LASSO, RF, and XGBoost), and evaluated by survival and ROC analyses. Immune infiltration, immune checkpoint expression, somatic mutations, and drug sensitivity were compared between risk groups. Single-cell RNA sequencing (scRNA-seq) data were used to map the distribution of prognostic genes across cell populations, and key genes and T cell infiltration were further validated by immunohistochemistry (IHC). Nine prognostic genes were identified in ESCC. Patients were divided into high- and low-risk groups based on the risk score, and the survival rate of the patients in the high-risk group was lower. Immune infiltration patterns, immune checkpoint expression, mutation landscapes, and drug sensitivity differed markedly between risk groups. scRNA-seq data analysis identified 11 major cell types and revealed distinct expression patterns of prognostic genes. IHC confirmed that FADD and SNAI2 protein levels were significantly upregulated in ESCC tissues, while a non-significant negative trend was observed between SNAI2 protein levels and T cell infiltration. These findings provide new insights into the molecular heterogeneity and immune microenvironment of ESCC, and suggest that the identified prognostic genes (CST2, FADD, SNAI2, MAGEA3, and SLCO1B1) may serve as candidate biomarkers for risk stratification. The observed negative trend between SNAI2 expression and T cell infiltration requires further validation in larger, well-matched cohorts.