Identification and validation of an anoikis-related gene signature based on machine learning algorithms in esophageal squamous cell carcinoma

Abstract

Abstract Purpose Anoikis plays a key role in the process of tumor metastasis. This study aims to investigate the characteristic of anoikis-related genes in esophageal squamous cell carcinoma (ESCC). Methods The differentially expressed anoikis-related genes in the TCGA-ESCC cohort were identified. The key anoikis-related genes were selected by least absolute shrinkage and selection operator (LASSO) regression and support vector machine-recursive feature elimination (SVM-RFE), and their expression were verified by qRT-PCR. Multivariate Cox regression was applied to construct the anoikis-related gene signature (ARGS). GSEA was applied to investigate the differences of biological function and pathway in different ARGS subgroups. Immune cell analysis was analyzed by ssGSEA. The expression characteristics of key anoikis-related genes in the single-cell dataset derived from the TISCH database. Immunotherapy response prediction was performed by the TIDE algorithm. Results The signature containing 5 key anoikis-related genes (CLDN1, EGFR, PLK1, SATB1, and TNFSF10) was constructed. CLDN1, PLK1, SATB1, and TNFSF10 were shown to be highly expressed in ESCC by qRT-PCR. The ARGS-high group had enriched in more abundant cells and immune-related pathways. Additionally, the ARGS-high group benefited well from immunotherapy, while the ARGS-low group was more sensitive to chemotherapy. Conclusion This study identified 5 key anoikis-related genes and conducted the ARGS, which can help predict prognosis and may guide treatment strategies for ESCC patients.