The Improvement Of Dynamic Bayesian Network Approach For Regulatory Networks

With the developed of the microarray technology.microarray date has increasingly be-come an important date source for bioinformatics research.The use of high-performancecomputer provide the powful tools for the research of gene regulatory networks. con-struction of gene regulatory netweoks with microarray is an important research topic offunctional genomics.Using BayesianNetworks to construct gene regulatory networks iscurrently the bioinformatics research focus. DBNs can incorporate the timing character-istics of date into models,can apply to many felds well(face identifcation,trafc and soon),estimating and forecasting for multiple moment state of the observation value. canexpress the feedback interaction of genes.It is ofer an efective tool for announcing generegulatory networks.This paper introduces basic theory of BayesianNetworks and DynamicBayesianNetworks, mainly discuss the principle of DynamicBayesianNetworks model as well asusing improved Bayesian Networks to construct gene regulatory networks:The innovation of this paper is: The missing data in building gene regulation networkis better dealt with SEM(Bayesian structre expectation maximization)algorithm,however,theresult of learning by SEM algorithm has strong depengence on the iniaial parameters.Thisdissertation proposed an improver particlefiltering algorithm,which randomly generateda number of candidate initial parameters and selectde the best parameters and selectdethe best parameter as whole model’s initial parameter to erecute basic SEM algorithmafter a iterative process.Comparing gene regulation network constructed with yeast cyclegene expression data by improved SEM algorithm with existing literature improve theaccuracy of constructing regulati on net work.