Identification Of Differentially Expressed Genes Between Subtypes Of Early Non-small Cell Lung Cancer Using Meta-analysis

Non-small cell lung cancer(NSCLC) is one kind of malignant tumors that poses a serious threat to human health. It includes two major pathological subtypes, they are lung adenocarcinoma(AC) and squamous cell lung carcinoma(SCC). Previous research has demonstrated that though both of these two diseases are kinds of lung cancer, there are fundamental differences in the basic carcinogenesis mechanisms, tumor growth and tissue invasion between them. However, the indiscriminate treatment of the two NSCLC subtypes has made unsatisfying curative effect for a long time. To treat these two major NSCLC subtypes with different therapeutic strategies, a understand of their differences on molecular level is required. Thus, the identification and investigation of differentially expressed genes(DEGs) between them is one of the most important means. In this paper, we applied two different meta-analysis to four microarray experiments to identify the DEGs between the two NSCLC subtypes. 164 AC and 161 SCC early-stage samples consisted of the whole sample set. Before meta-analysis,raw gene expression values were converted into two forms,they are the probability of expression(POE) representing gene’s differentially expressed probability, and expression barcode values representing its expression status(on/off) respectively. The results indicated that compared to POE meta-analysis, barcode meta-analysis can not only reduce the heterogeneity in genes across studies but also gain more effective information. DEG sets identified by two meta-analysis overlapped significantly(P=1.3x10-4) but the barcode meta-analysis yielded a lower global FDR. Consequently, based on this and other evaluated statistics of performance, we concluded that meta-analysis using barcode data outperformed POE method. Finally, we conducted ontology annotation and pathway analysis on those DEGs. The results indicate that many genes and the pathways they involved in were found closely related with different subtypes of NSCLC.