The human Drug–Disease–Gene Network

Abstract

How to understand the correlations of drugs, diseases and genes from a view of network-based framework is of great importance for disease diagnosis, treatments and drug discovery. This paper proposes a multi-level network model that integrates drugs, diseases and genes together, called a drug–disease–gene network (DrDiGeN). The network consists of three subnetworks, a drug–drug network (DrDrN), a disease–disease network (DiDiN) and a gene–gene network (GeGeN). The statistic characteristics show that node degree in most of the subnetworks approximately follows a power-law distribution. The results also indicate that if genes in the GeGeN occupy important topological positions, then their associated drugs and diseases always hold critical roles in the DrDrN and the DiDiN respectively. In addition, most drug target genes and disease-causing genes are always different and nonessential, and the both show a lower likelihood to encode hub proteins in human protein–protein interaction (PPI) network, while a little higher tendency is observed in the GeGeN. Gene modules extracted from the GeGeN are highly enriched in Gene Ontology (GO) terms, but poor coexpressed in human tissues compared with that of the PPI network. Furthermore, diseases (or drugs) associated with similar genes highly interact with each other such that tightly related drugs, diseases and genes can easily form co-modules, in which they share a similar pattern. The conserved structures are helpful for the understanding of the interaction mechanisms of drug–disease–gene as well as drug applications and disease treatments in a network-based level.