Screening of Demethylation-related Biomarkers and Exploration ofRegulatory Mechanisms in Esophageal Cancer Patients Based on MachineLearning and Mendelian Randomization

INTRODUCTION: The aggressive cancer known as Esophageal Squamous Cell Carcinoma (ESCC) has a dismal prognosis. Epigenetic changes such as demethylation have a significant impact on ESCC development and progression. METHODS: WGCNA and differential expression analysis of GEO datasets identified tumorrelated demethylation genes. Random Survival Forest (RSF), univariate Cox regression, and LASSO-Cox models were employed to further screen prognostic genes and establish risk prediction models. Drug sensitivity and immune infiltration analyses were used to evaluate therapeutic implications. Mendelian Randomization (MR) assessed the genetic causality of key genes. Single-cell RNA sequencing elucidated cellular heterogeneity, intercellular communication, and differentiation trajectories in the tumor microenvironment. Finally, qPCR validated key gene expression in both ESCC tumor tissues and adjacent normal tissues. RESULTS: Integrating co-expression with differential expression analyses enabled the identification of 150 demethylation genes associated with tumors. Six key prognostic genes (SERPINH1, PLAU, ANO1, RAB25, MAGEA4, and COL2A1) were selected to develop a risk prediction model, which showed improved accuracy after integrating clinical variables Age and Stage. Risk scores positively correlated with tumor stage and patient age, with higher scores predicting increased sensitivity to cisplatin, docetaxel, and vinorelbine. Immune infiltration analysis revealed reduced neutrophil levels associated with key gene expression. MR identified PLAU and RAB25 as causally linked to ESCC. Single-cell transcriptome and cell communication analyses highlighted squamous epithelial cells and altered LAMININ signaling in tumors. qPCR and expression analysis verified the expression of key genes. DISCUSSION: This study identifies demethylation-driven prognostic genes as possible targets for treatment and biomarkers for precision management of ESCC. CONCLUSION: Six prognostic genes, particularly RAB25 and PLAU, influence ESCC progression and immune microenvironment remodeling.