Research On Gene Regulatory Networks Construction Algorithms Based On Bayesian Networks

With the completion of Human Genome Project, the discovery of the gene function has become the new research interest in the Post-genome times. Using the microarray techniques, it is possible for scientists to discover the regulatory relationships between genes. The gene regulatory network, in definition, is the simulation or reconstruction of the mutual relations among expressed genes. Gene regulatory network helps us to understand in organisms which, where, when and how genes are expressed though observing visual model of gene expression. In this way, gene regulatory network has been widely applied in the research on relations between genes and diseases or drug target designs.The algorithms of reconstructing gene regulatory networks based on Bayesian network are the main research interest of this paper. Firstly, we study some models which are often used in reconstructing gene regulatory networks, weighed matrices model, Boolean networks model, mutual-information networks model, linear combination model, directed graphs and undirected graphs, Bayesian networks model and analyze the strengths and weakness of each model by discussing each model's details. As an effective measure, Bayesian network model is good at dealing with hide variables and absent values. We focus on devising the Bayesian network learning algorithm in order to establish an accurate prediction model. In case of the feature of gene expression data, it is not proper to directly deal with gene expression data by usual machine learning algorithms and mutual information method are used to preprocess the data in advance. Further, we present a sparse candidate algorithm based on mutual information and combine it with other reconstruction algorithms. As a result, two improved algorithms, CP (Conditional Independent Parent) and MP (Markov blanket Parent) have been realized. Experiments show that our algorithms were better than simulated annealing algorithm or other heuristic searching algorithms. Finally, we make the networks visualized.