Text mining and network analysis of molecular interaction in non-small cell lung cancer by using natural language processing.

Lung cancer including non-small cell lung cancer (NSCLC) and small cell lung cancer is one of the most aggressive tumors with high incidence and low survival rate. The typical NSCLC patients account for 80-85 % of the total lung cancer patients. To systemically explore the molecular mechanisms of NSCLC, we performed a molecular network analysis between human and mouse to identify key genes (pathways) involved in the occurrence of NSCLC. We automatically extracted the human-to-mouse orthologous interactions using the GeneWays system by natural language processing and further constructed molecular (gene and its products) networks by mapping the human-to-mouse interactions to NSCLC-related mammalian phenotypes, followed by module analysis using ClusterONE of Cytoscape and pathway enrichment analysis using the database for annotation, visualization and integrated discovery (DAVID) successively. A total of 70 genes were proven to be related to the mammalian phenotypes of NSCLC, and seven genes (ATAD5, BECN1, CDKN2A, FNTB, E2F1, KRAS and PTEN) were found to have a bearing on more than one mammalian phenotype (MP) each. Four network clusters centered by four genes thyroglobulin (TG), neurofibromatosis type-1 (NF1 ), neurofibromatosis type 2 (NF2 ) and E2F transcription factor 1 (E2F1) were generated. Genes in the four network modules were enriched in eight KEGG pathways (p value < 0.05), including pathways in cancer, small cell lung cancer, cell cycle and p53 signaling pathway. Genes p53 and E2F1 may play important roles in NSCLC occurrence, and thus can be considered as therapeutic targets for NSCLC.