Prediction of Overall Survival in Gastric Cancer by a Six-Gene Cox Proportional Hazards Model

This study aims to investigate prognostic genes that are correlated with overall survival in gastric cancer. A list of 43 critical genes is first selected from literature review and is narrowed down using marginal analysis, false discovery rate (FDR). This study used the 16 differential genes selected by FDR to create a cox proportional hazards regression model. Using a stepwise approach, the cox model is refined. From the 16 genes, 6 genes significantly correlated with overall survival are chosen and kept in the cox model. A principal component regression model was also constructed based on the principal component analysis results. The concordance of the principal component regression model is then compared to the concordance of the cox proportional hazards regression model. The final 6 identified genes are NRP1, STK11, MCM2, MARCKS, CTS6, C5. Of the 6 genes, MARCKS, NRP1, STK11, MCM2 are in line with previous research. CTS6 and C5, although studied in other cancers, are comparatively novel in the field of gastric cancer