Background: Despite advances in colon cancer screening, diagnosis, chemotherapy, and targeted therapy, the prognosis remains poor once colon cancer develops distant metastasis or local recurrence. To further improve the prognosis of colon cancer patients, researchers or clinicians may need to identify new indicators for predicting the prognosis and treatment of colon cancer.
Methods: In order to discover the new mechanism of epithelial-mesenchymal transition (EMT) promoting tumor progression and to find new indicators of colon cancer diagnosis, targeted therapy and prognosis, this study conducted The Cancer Genome Atlas (TCGA) analysis, differential gene analysis, prognostic analysis, protein-protein interaction (PPI), enrichment analysis, molecular typing, and a machine algorithm were combined with data from TCGA and Gene Expression Omnibus (GEO) databases and EMT-related genes.
Results: Our study identified 22 EMT-related genes with clinical prognostic value in colon cancer. On the basis of 22 EMT-related genes, we divided colon cancer into 2 different molecular subtypes by non-negative matrix factorization (NMF) model using 14 differentially expressed genes (DEGs), and the DEGs were enriched in multiple signaling pathways related to tumor metastasis process. Further analysis of EMT DEGs revealed that the PCOLCE2 and CXCL1 genes were characteristic genes for clinical prognosis of colon cancer.
Conclusions: In this study, 22 prognostic genes were screened out from 200 EMT-related genes, and then the PCOLCE2 and CXCL1 molecules were finally focused on through the combination of the NMF molecular typing model and machine learning screening feature genes, suggesting that PCOLCE2 and CXCL1 may have good application potential. The findings provide a theoretical basis for the next clinical transformation in the treatment of colon cancer.
Keywords: CXCL1; Epithelial-mesenchymal transition (EMT); PCOLCE2; molecular typing; random forest algorithm.
2023 Journal of Gastrointestinal Oncology. All rights reserved.