Macrophage-related Genes were Discovered by Applying Highdimensional Weighted Gene Co-expression Network Analysis asBiomarkers for Esophageal Squamous Cell Carcinoma

Introduction: Esophageal squamous cell carcinoma (ESCC) lacks effective early diagnostic markers. High-dimensional WGCNA (hdWGCNA) enables the systematic identification of macrophage-related signatures from single-cell data, offering a novel approach to discovering potential biomarkers for ESCC. Methods: This study analyzed bulk transcriptomic datasets and a single‑cell RNA‑seq dataset from the Gene Expression Omnibus (GEO) database. Specifically, the GSE53625 dataset (179 ESCC and 179 matched normal samples) was a training cohort, while GSE20347 (17 ESCC and 17 normal samples) was an independent validation set. The single‑cell dataset GSE196756 (3 ESCC and 3 adjacent normal tissues) was processed to construct a cellular atlas. The CellChat package was used to assess the communication between macrophages and other cell subpopulations. Macrophage‑associated co‑expression modules were identified by hdWGCNA. Differentially expressed genes (DEGs) were screened with the limma package, and the intersection of DEGs and hub genes from hdWGCNA was further refined via LASSO and SVM-RFE algorithms to obtain candidate genes. Functional enrichment, immune infiltration, drug prediction, and molecular docking were subsequently performed. In vitro experiments were carried out to validate the expression and functional role of candidate genes in ESCC cell lines. Results: Four macrophage‑related gene models consisting of TGFBI, FTL, GPNMB, and APOE were created and validated. These genes were significantly overexpressed in ESCC tissues and exhibited high diagnostic accuracy, reaching an AUC value >0.8 in both the training and independent validation cohorts. In vitro functional assays confirmed that silencing TGFBI suppressed ESCC cell proliferation, migration, and invasion. Enrichment analysis linked these genes to neutrophil activation and apoptosis pathways, and immune infiltration analysis further revealed their correlation with immunosuppressive cell subsets. Drug prediction yielded 190 candidate compounds, and molecular docking suggested potential binding of retinoic acid, rosiglitazone, and GDC‑0941 to the corresponding targets in ESCC. Discussion: In this study, four characteristic genes (TGFBI, FTL, GPNMB, and APOE) were identified for the diagnosis of ESCC, contributing to the improvement of early detection of the cancer. Conclusion: The present findings contributed to the pathogenesis research, detection, and management of ESCC.