Abstract: The aim of the present study is to investigate the pathway of key genes that participate in the development of lung cancer and construct the prediction model. Firstly, the differentially expressed genes(DEGs) between cancer and healthy group were identified by analyzing the transcription profiles of GSE19188 and GSE40791. The two transcription profiles were downloaded from the GEO(Gene Expression Omnibus) database. Secondly, the DEGs were entered into the STRING database to construct the protein-protein interaction(PPI) network, and the key genes were screened by the product of expression deviation score and degree in the network. A total of 805 DEGs in the same trend were screened in lung cancer samples compared to normal tissue samples, including 214 up-regulated and 591 down-regulated genes. The PPI network of DEGs comprised 632 DEGs and 6312 interaction pairs. A total of 209 key genes were obtained, which were significantly enriched in seven up-regulated pathways and four down-regulated pathways. Hierarchy analysis discovered that the pathscore of the 11 significant pathways could distinguish cancer samples from normal samples. Finally, cross talk genes in the connected pathways were analyzed and used to construct the prediction model through the machine learning method. The accuracy of the prediction model could be as high as 97%, indicating that the 18 genes had good robustness and sensitivity as the prediction genes for lung cancer. So, these genes may be used as the prediction markers and target biomarkers for chemotherapies in lung cancer.

Key words: lung cancer, differentially expressed genes(DEGs), protein-protein interaction(PPI) network, cross talk gene, prediction model