Improving cancer driver genes identifying based on graph embedding hypergraph and hierarchical synergy dominance model

Cancer driver genes play a crucial role in the formation and development of cancer. Their mutations provide tumor cells with a selective growth advantage, thus promoting tumor development. Identifying these genes is essential for understanding cancer pathogenesis at the molecular level and developing targeted drugs. In this paper, a new method for identifying cancer driver genes, named GHSDM, which is based on graph embedding hypergraph and hierarchical synergy dominance model, is proposed. Firstly, a global non-binary mutation matrix is constructed using gene expression data and somatic mutation data. Secondly, global non-binary mutation matrix is combined with the PPI network using Node2vec and K-means algorithms to construct a hypergraph. From this hypergraph, the entropy clustering coefficient (EntroCC) is extracted. Additionally, gene expression data are integrated with DNA methylation data to obtain a new feature score, EDES. Finally, the hierarchical synergy dominance model (HSDM) is proposed by analyzing the correlation among features. HSDM is then used to fuse EntroCC, EDES,and miRNA importance score. To validate the performance of GHSDM, experiments are conducted on four real cancer datasets. The results show that GHSDM outperforms six current classical methods in statistical evaluation metrics, functional consistency, and the partial area under the ROC curve (PAUC), demonstrates robust cross-cancer capability.