Development of a Machine Learning-Based Programmed Cell Death Genes Prognostic Model for Esophageal Carcinoma

Background This study aimed to evaluate the prognostic value of Programmed Cell Death Related Genes (PCDRGs) in esophageal carcinoma (EC) using bioinformatics methods, and to explore EC development and immune regulatory mechanisms from multiple perspectives. Methods RNA-seq expression and clinical data of EC patients were obtained from the TCGA and GSE53622 datasets, along with 18 key regulatory PCD pattern genes. A predictive model, termed ‘Characteristics of Cell Deaths’ (CDS) was developed by integrating 10 different machine learning methods. The model identidied seven prognosis-related genes were screened out, whose association with EC was analyzed. Results The PCDRGs-based prognostic model constructed using the StepCox[both] + Random Survival Forest (RSF) method demonstrated optimal performance. CDS exhibited robust and significant predictive capability for EC clinical outcomes and was validated as an independent risk factor in both TCGA and GEO datasets. Conclusion This study successfully developed a novel EC PCDRGs-based prognostic model for EC. With further validation, this model has the potential to predict patient prognosis and sensitivity to drug treatments, contributing to personalized therapeutic strategies in EC management.