Reconstruction And Statistical Analysis Of Human Disease Network

Human Disease Network has become a powerful tool for discovering the relationship between disease and disease,disease and genes.However,the number of currently iden-tified disease genes is very limited.The correlation among disease genes is an important research topic.Construction of related networks on the basis of the data from the human online Mendelian Inheritance in Man(OMIM)database in December 2016 and the paper published by Kwang-11 Goh et al.in 2007 in PNAS is interesting yet important.The data are merged according to disease types and disease phenotypes,which include 26 diseases categories,3705 phenotype categories,8233 disease phenotypes and 5100 different genes.Then,based on the obtained data,a new human disease network(nHDN)and a new disease gene network(nDGN)are constructed by using bipartite graphs.The following conclusions can be drawn from the analysis of network topology,modularity analysis and functional analysis.First of all,topological analysis of the constructed networks show that the degree distributions of nHDN,nDGN and their largest connected subgraphs approximately follow power-law distributions,indicating that the constructed networks are all close to scale-free networks;by computing assortativity coefficients.We found that the four networks are assortative,that is,large degree nodes tend to connect with the large degree ones.Our research also found that the largest connected subgraph of the human disease network has small-world characteristics.Secondly,modularity analysis of the nHDN was performed by calculating the levels of heterophilicity(H)and dyadicity(D).It is found that diseases tend to gather into tight communities,and genes related to the same disease category have a certain correlation in function,and the relationship among different disease categories is quite weak.But the chromosome diseases are less modular and have weaker associations with other diseases.The largest connected subgraphs are randomized by degree-preserving and degree-distribution-preserving and then we compared their statistical properties with the actual networks.It is found that in the actual network,diseases(genes)in the same disease category are more likely to be linked with each other.Functional analysis of nHDN and nDGN networks revealed that genes associated with the same disease phenotype are enriched in the same or similar GO biological processes,molecular functions,and KEGG pathways.