| In the post-genomic era, one of the fundamental challenges in the biological field remains understanding the complex gene regulatory networks that control cellular functions. With the development of high-throughput microarray technologies, it brings a great deal of microarray data to life science research, and constructing gene regulatory networks from these experimental data makes cellular functional research possible at the molecular level. Because the expression of one gene will be influenced by other genes, and the gene will also influence the expressions of other genes, these interactions will construct a complex gene regulatory network. So gene regulatory networks describe the relationships among many genes which complete different tasks cooperatively. Constructing gene regulatory networks can make us do some whole and simulating analysis and researches for the expression relationships of all genes of a certain species or tissue, so that we can know life phenomena and reveal the fundamental laws of life activities.Currently, Bayesian network methods are used to construct gene regulatory networks extensively. However, some structure learning algorithms of Bayesian networks need known node ordering for structure learning, such as K2 and TPDA-∏algorithm. But in the absence of prior information, it is impossible to gain node ordering, while the accuracy is quite terrible by using stochastic node ordering. To a large extent, it limits the application of structure learning methods of Bayesian networks. In this paper, we propose a novel method of inferring node topologic ordering, which makes up for the deficiency of Bayesian networks needing known node ordering. Then through combining it with TPDA-∏algorithm and K2 algorithm respectively, we form two new algorithms inferring structure of Bayesian networks called NO-TPDA-∏and NO-K2. Finally based on the new method, we did some simulation experiments using benchmark datasets and yeast gene chip datasets, and obtained great results. |
PDF Full Text Request