J Dig Dis. 2020 Sep 8. doi: 10.1111/1751-2980.12936. Online ahead of print.
OBJECTIVE: Liver and lung metastasis are the most observed distant metastasis of colorectal cancer. How primary colorectal cancer cells choose to colonize to liver or to lung is not clearly elucidated. Here, we aimed to identify key genes and pathways that are capable of predicting distant metastasis of colorectal cancer to liver or to lung.
METHODS: 3 gene expression array datasets from Gene Expression Omnibus (GEO) were analyzed. Protein-protein network analysis, best subsets regression and backward stepwise regression were used to screen key genes. Expression of key genes were used to construct a predicting logistic regression model. The expression data from local clinical samples were used as validation dataset. The Receiver operating characteristic curve (ROC) was used for testing the performance of the predicting model.
RESULTS: 59 liver-metastasis-related differentially expressed genes (DEGs), 90 lung-metastasis-related DEGs and 45 liver/lung-metastasis-shared DEGs were identified. KEGG pathways and GO terms that enriched in liver and lung metastasis were recognized. A predicting logistic regression model consisted of SPARC, COL1A2, MMP9, COL11A1, COL3A1, CXCL12 and THBS2 were established. ROC of the model reached an 83.9% area under the curve (AUC) in predicting liver and lung metastasis by using our clinical samples as validation dataset.
CONCLUSION: 7 key genes capable of predicting liver/lung metastasis of colorectal cancer were identified, which would make clues for exploring the mechanism of the target organ selection during metastatic process in colorectal cancer and inspire new potential targets for metastasis inhibiting therapy. This article is protected by copyright. All rights reserved.