ABSTRACT
Background Colorectal cancer (CRC) is the most common malignant tumor of digestive system. The metastasesis the main cause of mortality in CRC patients, of whom the initial diagnosis is about 25%. In our study, weaimed to identify potential gene biomarkers based on RNA sequencing data to predict and improve CRCpatient survival.Method In this study, by screening differentially expressed genes of colon cancer related to liver metastasis, asurvival prognostic risk model was constructed by bioinformatics analysis. Here, we conducted our data mininganalysis for CRC by integrating the differentially expressed genes acquired from Gene Expression Omnibus(GEO) database by primary tumor versus liver metastasis (GSE81582，GSE41258，GSE49355，GSE68468)into The Cancer Genome Atlas (TCGA) database which includes 415 primary tumor and 132 liver metastasistissue. At the same time, we used transwell, RT-PCR and western to examine the effects of CLCA1 and SPINK4on the migration of colorectal cancer cells at the cell level.Results We identified intersections of 197 genes (117 up-regulated and 80 down-regulated) between GEO dataand TCGA data. Differentially expressed genes in TCGA-COAD by single factor cox analysis, lasso cycle trainingand multifactor cox analysis composed a survival prognosis prediction model consisted of 7 genes ORM1,CLCA1, C8B, SPINK4, ALDOB, GAMT, C8G. And results of transwell experiments showed that high expression ofCLCA1 and SPINK4 can inhibit the migration ability of colon cancer cells LOVO and SW620, meanwhile westernblotting showed that the high expression of both genes can upregulate the expression of epithelial phenotypicmarker E-cadherin, and Vimentin expression is down-regulated.Conclusion In this study, 197 differentially expressed genes were selected and a relatively robust survivalprognosis prediction model was constructed. The model consisted of seven genes: GAMT, C8G, ORM1, CLCA1,C8B, SPINK4, and ALDOB. At the same time, we found that CLCA1 and SPINK4 are closely related to survivalprognosis. The predictive model nomogram will enable patients with CRC to be more accurately managed intrials testing new drugs and in clinical practice.