Abstract.We have identified an 8-gene signature with significant expression differences between gastric cancer and normal gastric tissues. This 8-gene set can predict the normal and cancer status of gastric tissues with more than 96% accuracy in a totally independent microarray dataset. The 8 genes are composed of down-regulated KLF4, GPX3, SST and LIPF, together with up-regulated SERPINH1, THY1 and INHBA in gastric cancer. To corroborate the differential gene expression pattern, we chose GPX3 and examined its expression pattern in detail. A comparison of GPX3 expression pattern shows a broader down-regulated pattern in multiple types of cancers, including cervical, thyroid, head and neck, lung cancers and melanoma than in healthy controls. An immuno-histostaining analysis in tissue microarrays confirms GPX3 down-regulation in gastric cancer. Mechanism-wise GPX3 down-regulation in gastric cancer is due to promoter hypermethylation. Collectively, these results show a correct identification of 8 genes as gastric cancer biomarkers.