一种改进的基于节点拓扑排序的基因调控网络构建方法

urrently some structure learning algorithms of Bayesian networks for constructing gene regulatory networks donˇt consider identifying the directions of the edges between the same node pairs in different sub-graphs. However, it is the common phenomenon in gene regulatory networks. For the shortage, we propose to use discrimination information for identifying the directions of the edges between the same node pairs in different sub-graphs, and establish a novel effective algorithm of inferring node topologic ordering from the datasets without prior information. Through combining it with K2 algorithm, a novel improved structure learning method, which makes up for the deficiency that Bayesian networks need the known node ordering , is formed and is suitable for the construction of gene regulatory networks. Two benchmark datasets with known network structures 僴were used to evaluate the novel method, and the results show that inferring network structures are close to the optimal ones, and prediction accuracy can be compared with the similar methods. Moreover its computational time and prediction accuracy are greatly superior to the Hill climbing method. When the novel method was applied to gene expression datasets, the obtained results can be better supported by the existing biological experimental evidence, which verified the effectiveness and feasibility of our method.