A machine learning-based gene signature reveals radiotherapy-induced remodeling of the immune microenvironment in esophageal cancer

This study aimed to develop a machine learning-based signature using radiotherapy-responsive genes to characterize immune infiltration dynamics and evaluate drug sensitivity in esophageal cancer (EC) before and after radiotherapy. Based on the GSE137867 dataset from GEO, which includes 4 pre-radiotherapy and 4 post-radiotherapy esophageal cancer samples, we identified significantly differentially expressed genes (DEGs) through differential analysis. Gene Ontology (GO) analysis and Gene Set Enrichment Analysis (GSEA) were performed based on the DEGs. Signature genes were screened using Lasso and SVM machine learning algorithms. These signature genes underwent CIBERSORT immune infiltration analysis, ssGSEA, ESTIMATE immune scoring, drug sensitivity analysis, and ceRNA network analysis. To validate the expression profiles of the target genes, we cultured the esophageal cancer OE19 cell line and performed Reverse Transcription Quantitative Real-Time Polymerase Chain Reaction (RT-qPCR). After identifying 28 DEGs, GO and GSEA analysis revealed that these DEGs were enriched in regulation of memory/regulatory T cell differentiation, chemokine activity and cytokine-cytokine receptor interaction pathways. Through Lasso and SVM, three signature genes were identified: ZDHHC11B, CD46, and PFKFB3. CIBERSORT analysis showed a marked elevation in CD8 + T cells after radiotherapy. Drug sensitivity analysis revealed that CD46 was negatively correlated with cisplatin, positively correlated with sapitinib, while PFKFB3 showed a negative correlation with linsitinib. The ceRNA network included 21 miRNAs, 3 mRNAs, and 18 lncRNAs. Specifically, 21 miRNA-mRNA regulatory pairs and 18 lncRNA-mRNA regulatory pairs were identified. RT-qPCR showed that compared with the control group, ZDHHC11B decreased significantly in the post-radiotherapy group, while CD46 and PFKFB3 increased significantly. This study comprehensively investigated the immune landscape of EC pre- and post-radiotherapy and analyzed drug sensitivity of these signature genes, thereby elucidating alterations in the immune microenvironment of EC and nominating ZDHHC11, CD46, and PFKFB3 as potential biomarkers for predicting immune status and therapeutic sensitivity.