Discovery and validation of prognostic markers in gastric cancer by genome-wide expression profiling

AIM: To develop a prognostic gene set that can predict patient overall survival status based on the whole genome expression analysis. METHODS: Using Illumina HumanWG-6 BeadChip followed by semi-supervised analysis, we analyzed the expression of 47,296 transcripts in two batches of gastric cancer patients who underwent surgical resection. Thirty-nine samples in the first batch were used as the training set to discover candidate markers correlated to overall survival, and thirty-three samples in the second batch were used for validation. RESULTS: A panel of ten genes were identified as prognostic marker in the first batch samples and classified patients into a low- and a high-risk group with significantly different survival times (P = 0.000047). This prognostic marker was then verified in an independent validation sample batch (P = 0.0009). By comparing with the traditional Tumor-node-metastasis (TNM) staging system, this ten-gene prognostic marker showed consistent prognosis results. It was the only independent prognostic value by multivariate Cox regression analysis (P = 0.007). Interestingly, six of these ten genes are ribosomal proteins, suggesting a possible association between the deregulation of ribosome related gene expression and the poor prognosis. CONCLUSION: A ten-gene marker correlated with overall prognosis, including 6 ribosomal proteins, was identified and verified, which may complement the predictive value of TNM staging system.