Disease Progression Analysis Based On Molecular Networks

Complex disease has a serious impact on human health and their normal life. As a complex disease with severe injury to human health, cancer has attracted lots of attention and it has been researched for a long time. Research shows that cancer has become a major cause that led to the world’s population death. With the rapid development of post-genome era and the technology of gene sequencing, researchers obtain more knowledge of genetic material and gene sequences. The increasing of sequencing data leads to more public databases, including TCGA, OMIM and GEO. These public data can be used for disease research, making studying of the potential mechanism and pathogenesis of cancer on the molecular level becoming an import direction of human disease research. On the other hand, the occurrence and development of diseases is a dynamic biological process. Disease changes a lot with time, which leads the research of complex disease on dynamic molecular networks attracting more attention. Evolution analysis of disease on dynamic networks can help understanding the function of gene and protein, exploring changes of biological pathways and judging the state of cells from the micro aspect. Also, it can help revealing the mechanism of disease, determining the efficacy of the drugs and formulating individualized treatment from the macroscopic aspect.Different from existed biomarker identification methods on static or dynamic networks, this paper aims to analyze the progression of complex diseases from the dynamic perspective. In this paper, we constructed the dynamic disease networks for the three phrases of lung cancer on the molecular level and analyzed the evolution of the disease on these dynamic networks. Combining the protein-protein interaction(PPI) network and the GEO(Gene Expression Omnibus) datasets, we built three molecular networks of the three phases of lung cancer and then detected module biomarkers on these dynamic networks using Markov clustering method(MCL) and built a Multi partite graph of the network evolution process. After that, we analyzed the evolution of lung cancer on the three networks from import nodes, pathways and modules respectively. The analyses include the size of networks, the distribution of degree and clustering coefficient of genes in networks, the related pathways to lung of each phases, the number of modules and their changes and evolutionary events. Finally, we scored eachgene in the networks depending on their relationship between known disease genes and the evolution pattern of disease and then sorted all genes. We selected genes with high scores as the predict candidate disease gene set and validated the predict genes from both statistical significance and biological significance.Evolution analyses on the dynamic disease networks depending on PPI network and microarray expression data(ID: GDS3257) show that our method to construct disease networks works well and the networks we built can reflect the interactions between genes. The evolution analyses of disease on the dynamic networks can illustrate the important changes of genes in the development process of lung cancer from the molecular level and identify related module biomarkers and pathways. The prediction and validation of candidate disease genes impacts the fact that the result of our method is not randomly selected and can be used as robust module biomarkers.