Gene Regulatory Network Building And The Analysis Of Dynamic And Stability

The building of gene regulatory network is a very important issue in biology. Recently,as the development of biological technologies, especially the rapid development of microar-ray technology, many gene expression profiles which described genome expression behaviorunder a specific life stage were supported. The expression value from gene expression pro-files can be treated as the posterior observation from gene regulatory network. Many differ-ent computational methods and models were suggested to rebuild gene regulatory networkstructure from gene expression profiles.The regulation from transcription factors to targets were studied as the first step. Wedescribed the method to improve the prediction accuracy with enhanced Bayesian classifierwhich considered the interactions of genes between neighbored time points. We also sug-gested using the activity of transcription factors to control the quality of gene expressionprofiles, and the discrete feature encoding to reduce the noise.When prior information of transcription factor-targets pairs is not enough to build aneffective training set, model which can rebuild gene regulatory network from gene expres-sion profiles directly will be required. We suggested a method which made use of knowntranscription factors, and combined Bayesian network model and genetic algorithm to learnout the network structure in nearly linear time. We also considered the correlations betweentranscription factors and GO terms, and suggested a novel method to get the regulatoryrelationships from transcription factors to GO terms.Real gene regulatory network is dynamic, while the models mentioned above are staticmodels. To model the dynamic of gene regulatory network better, we used Boolean networkwhich is one of the most used dynamic model. The states transition and attractors in Booleannetwork are nice descriptions of network dynamic. When Boolean network was applied toan embryo heart development, the dynamic of Boolean network matched expected genebehaviors very well. To real gene regulatory networks, the dynamic of network should be kept under pertur-bations. We simulated the dynamic behavior of Boolean networks under kinds of perturba-tions to get the stability of network. Furthermore, we considered the statistical significanceof the dynamic of a network under purely random networks backgrounds.Finally, we supported two R packages, which can be used on gene regulatory networkbuilding and network dynamic analysis.