Identification and Validation of an Anoikis-Related Gene Signature to Predict Prognosis in Colorectal Cancer

Abstract

Abstract Purpose Colorectal cancer (CRC) is highly aggressive, with advanced tumors resulting in poor prognosis. Anoikis is a type of programmed cell death that is important in malignant solid tumor occurrence and progression. However, research on the role of anoikis in CRC and its prognosis is lacking. Methods Using patient data from the Gene Expression Omnibus (GEO) and The Cancer Genome Atlas (TCGA), anoikis related genes (ARGs) were identified. Gene Ontology (GO) and Gene Set Enrichment Analysis (GSEA) were performed to explore ARG-related biological pathways. We used least absolute shrinkage and selection operator (LASSO) regression analysis to construct a prognostic model, and the LASSO-derived median risk score could divide the cancer group into high- and low-risk groups. The prognostic value of ARGs was analyzed using multivariate COX regression and receiver operating characteristic curves (ROCs). We used decision curve analysis (DCA) to evaluate the clinical utility of the constructed prognostic model. Results We identified 21 differentially expressed genes, and the GO and GSEA analyses showed that genes in the dataset TCGA-COADREAD were significantly enriched in the WNT signaling pathway and pluripotency, negative regulation of NOTCH4 signaling, PI3K-AKT signaling pathway, and L1CAM interactions. Eight genes were verified in the GSE17536 and TCGA-COADREAD datasets (BRCA2, CXCL8, ITGA2, KLF4, PLAU, SOX9, TPM1, VSNL1). DCA indicated that the model's 5-year predictive effect was better than that at 1 and 3 years. Conclusions We demonstrated the value of ARGs to assess CRC prognosis, potentially providing new insights into CRC survival prediction and therapeutic targets.