Characterization Of Cancer Driver Genes Based On Network Method

According to global cancer statistics issued by the World Health Organization in 2018,cancer is one of the leading causes of death worldwide.Precision medicine would be extremely promoted by targetting the cancer driver genes with a key role in the tumorigenesis.In this study,four cancer consensus driver genes,including lung cancer,breast cancer,prostate cancer and colorectal cancer,were collected to study their biological functional characteristics.Then,the topological characteristics of those driver genes were characterized by different ways in original PPI network and cancer-specific differentially expressed genes and their neighbor(DEGN)PPI networks.Based on these characteristics,a cancer driver gene classification model was constructed and combined with the corresponding cancer somatic variation data to predict cancer driver genes.It was found that there was a great heterogeneity among the four cancer driver genes.GO enrichment analysis showed that they were significantly enriched in gland development related to tumor invasion and metastasis,while KEGG pathway enrichment analysis showed that they were all enriched in cancer-related items.Besides,the distribution characteristics of the driver genes in the two PPI networks shown that they were significantly enriched in the top 10%region with the highest centrality,the smallest 10% region of the average shortest path,and the 60%to 80%region with lower clustering coefficient and topological coefficient.This reflects that cancer driver genes have the core role in these two networks,and their role is difficult to be replaced by other nodes of the networks that indirectly emphasizing the important functional role of cancer driver genes.Finally,we integrate these characteristics to construct a cancer driver gene classification model and combined with the corresponding cancer somatic variation data to predict the potential driver genes.And it has better performance is than the current two popular network-based algorithms in lung cancer,breast cancer and colorectal cancer.Our study collates the most comprehensive,standardized and high confidence consensus driver genes for these four cancers,which is of great significance for driver gene recognition algorithms to verify their performance.Furthermore,this study systematically characterizes the characteristics of cancer driver genes and then uses these characteristics to construct models and achieve better performance.This provides good practice for systematically understanding the driver gene and improving the performance of the driver gene recognition algorithm.