The Research Of Learning Gene Regulatory Network Based On Technical Of Bayesian Network

The availability of different types of high-throughput experimental data, especially gene chips have greatly impelled the research of functional genomics. Because of the complexity of gene express data and the shortage of bioinformatics knowledge,here are still no relatively mature methods for varieties of gene expression mining analysis.Because of the abilities to handle uncertainty and represent dependency relationships,Bayesian Networks,a kind of statistics graph models,are now the focus for a growing number of researchers concerned with discovering novel interactions,information dependencies and gene regulatory networks from gene expression data.Bayesian network methods have shown promise in gene regulatory network reconstruction.The main contents of this dissertation are as follows:(1)A survey about the research on Bayesian networks was made,including the background,the current research state and development trend of Bayesian networks, the basic principle of Bayesian networks,the introduction and analysis of the classic,algorithms for inferring and learning of Bayesian networks.(2)K2 algorithm,which is the one of the most important method in Bayesian structure learning,is effective when the input nodes have been correct ordered. However,in most cases,this algorithm is applied without giving special attention to this preorder.An effective gene regulatory network construction method,which was proposed,named IE_K2 algorithm,firstly it constructs an undirected network based on mutual information between two nodes,then the union information entropy was introduced to determines the direction of each edge,the best order of nodes would be obtained as the input of the K2 Algorithm.The method is evaluated using alarm network.The results show that the proposed method can identify networks which are approximate to the optimal structures.It outperforms hill climbing methods and the random ordered K2 algorithm in terms of predicted structure accuracy.At last,IE_K2 algorithm is applied to the yeast cycle gene expression data.The effectiveness of the proposed method is illustrated by verifying a part of the inferred regulations through existing literatures.(3)Most research work in learning gene networks assumes that either there is no time delay in gene expression or there is a constant time delay.However,the biological literature shows that different gene pairs have different time delays for gene regulation.An effective gene regulatory network construction method,named TD_GN algorithm,is proposed.Firstly,it chooses candidate parent sets for each gene,based on mutual information and union information entropy and KL divergency.Then the new learning algorithm based on modification,named LM algorithm,was proposed to learn gene network.The experiments on both artificial and yeast cycle gene expression datasets evaluate the effectiveness of TD_GN learning algorithm.The learning performance of TD_GN is significantly better than that of K2 and REVEAL and DBmcmc.