Integrative analysis and validation of robust gene signature in lung cancer

Abstract

We studied robust gene signature (RGS) in lung cancer by using an approach of integrating a highly diverse collection of cancer genome-wide datasets, which were six public microarray datasets, one pair of Suppression Subtractive Hybridization EST library, one pair of Serial Analysis of Gene Expression (SAGE) experiments, and 191 Loss of Heterozygosity (LOH) reports obtained from 388 publications. Among the 109 RGS genes identified from our study, 14 of the 15 reported differentially expressed genes (DEGs) based on literature verification were consistent with our predictions. Out of the remaining 94 genes that were not reported as DEGs in lung cancer by any publication, we randomly picked eight and verified their expression in lung cancer versus normal tissues by semi-quantitative RT-PCR amplification, and all showed consistent expression pattern with our findings. System assessment analysis revealed that our integrative method had an accuracy of 95% and a correlation coefficients value of 0.92.