| 其他摘要 | Many drug target proteins (DTPs) have been defined by previous studies, and the DTPs are mainly distributed in several druggable families. Although proteins in one druggable family are similar in sequence and structure, not all of them are DTPs. However, the difference between DTPs and non-target proteins (NTPs) in druggable families is not yet clear. We explored the difference of DTPs and NTPs from known druggable families in human protein interaction network, and they differed significantly in several topological features, such as degree. We also found the similarity of topological features, which quantified by Pearson correlation coefficient (PCC) of six features for each two proteins in druggable families, was different between DTPs and NTPs. Most PCCs in DTPs were higher than those in NTPs, but PCCs in DTPs were diverse for different druggable families. The PCCs of withdrawn DTPs were also less than those of experimental DTPs. At last we found DTPs influenced the network to a larger extent compared with NTPs, which suggest that the difference between DTPs and NTPs might caused by efficiency. Thus, the druggability of proteins might relate to protein’s position in the network, which could be helpful to find new DTP candidates. Since organism development and many critical cell biological processes are organized in a module pattern, many algorithms have been proposed to detect modules recently. But no work constructed a module-module communicating network to interpret how these processes interact with each other. A new method, MOfinder, was presented to detect overlapping modules in protein-protein interaction (PPI) network. It demonstrated that our method is not only more sensitive and simpler, but also more reliable than common algorithms. Then, MOfinder was applied to human PPI network, and 451 overlapping modules were found. Using these modules, the module-module communication network is constructed, and it comprised by two large clusters and tens of small clusters. Disease gene annotation shows that these large clusters are cancer related, and the immune related protein always include in cancer related modules which offer some clues for cancer therapy by targeting immune genes. Our study about modules and communicating proteins enables the analysis of protein–protein interaction networks in a new perspective and thus may benefit network research and improve our understanding of disease. Cardiovascular disease (CVD) and other complex diseases cannot be explained by the“one gene one phenotype” rule. The system-level analysis is required by identifying candidate genes related to complex diseases, uncovering how environment factors (ENF) increase the possibility of sick, and improve discovering novel drug target candidates. Thus the disease network, ENF-gene network had been built. The power-law distribution observed in many other biological networks has been found in the CVD network. None of the ENF associated hub genes holds the hub property in CVD network, which suggest that hub genes keep away from environment factors in order to maintain the robustness of cardiovascular system. Combine the two networks, we found several shared genes between them and test their influence to the network: they hardly changed the network topological features. Besides, the ENF associated genes which overlap with CVD network were suggested as drug targets because both ENF and disease gene can be inhibited. We thought such situation should be considered in filtering candidate drug targets for other complex disease. The network-based approach will be helpful for understanding the mechanisms of complex diseases and finding suitable drug targets. |
修改评论