ABSTRACT
Objective To identify biomarkers associated with the differentiated phenotype based on gene expression profiling of gastric cancer. Methods Two bioinformatic methods, BAGEL and k-TSP, were used to identify featured genes associated with differentiation in gastric cancer samples based on the Oligo gene chip data, and ROC curves were used to verify the classification sensitivity and specificity of the identified genes. Finally, a total of 30 gastric cancer samples with different differentiation levels were collected for laboratory validation using real-time PCR analyses. Results A total of 121 differentially expressed genes were identified using the BAGEL algorithm, the criterion were FC ＞ 2. 0 and P ＜ 0. 001.Then, the k-TSP algorithm for feature selection based on this differential expression data were used, and 3 groups of featured genes which had potential to classify poor and well differentiation gastric cancer samples were identified, including MYLIP and TMPRSS3, ZNF266 and TM4SF1, SNAI2 and CNFN. To define the featured gene groups that had the highest classification capability, ROC curves to calculate the classification sensitivity and specificity of each gene group were used. The results showed that the combination of SNAI2and CNFN as a classifier had the highest classification sensitivity and specificity. Real-time PCR results showed that 18 of 22 poor differentiation samples were found with high expression of SNAI2 and low expression of CNFN (82%); 6 of 8 well differentiation samples were of low expression of SNAI2 and high expression of CNFN (75%). Conclusion The results indicate that SNAI2 and CNFN are constantly expressed in poor or well differentiation gastric cancer samples, and the expression pattern of these two genes is opposite. These results indicate that SNAI2 and CNFN have the potential for the identification of the differentiation level of gastric cancer.