Research On Pathogenic Genes Identification Method Based On Structural Controllability Of Complex Network

Complex networks refer to a class of networks with complex topologies abstracted from real complex systems,such as social networks,transportation networks,biological networks,and so on.In recent years,with the development and maturity of control theory,the study of structural controllability of complex networks has become a hot topic.Studies have shown that the structural controllability analysis of complex networks has practical application value in identifying potential pathogenic genes and drug targets.Based on the structural controllability model of complex network,we proposed a new controllability node classification framework and applied it to human tissue-specific regulation networks to systematically detect pathogenic genes with significant biological significance,providing a tool platform for further disease diagnosis and treatment.The main contents of this thesis are as follows:Firstly,a new controllability node classification framework was proposed based on structural controllability model.For complex network,different classification methods can be applied to network nodes from different perspectives of control:(1)node classification method based on controllability;(2)node classification method based on control capability;(3)node classification method based on control function source;(4)node classification method based on control edge robustness.Accordingly,a four-dimensional vector was built in which each component represented a node classification method.Secondly,candidate disease genes were discovered.For 32 human tissue-specific regulatory networks,the controllability node classification framework was applied and the genes were divided into different types.Then,statistically significant analysis and verification was performed on each type of genes on gold standard data sets with known functions,and a new type of biologically significant genes was identified which was defined as candidate disease genes.Furthermore,according to the correspondence between tissues and diseases,the disease specific genes on the tissue were screened out from the identified candidate disease genes.By analyzing the tissue difference of disease specific genes and literature mining and GO terms verification,it was shown that the selected disease specific genes are indeed closely related to the tissue diseases.Based on the structural controllability model of complex network,a new controllability node classification framework was proposed.We applied it to thirty-two tissue regulatory networks to find out the biologically significant gene types,i.e.candidate disease genes,and further analyzed the correlation between candidate disease genes and diseases.The experimental results showed that the controllability node classification framework based on the controllability model of complex network can help identify disease-related genes.