摘要
动态贝叶斯网络(dynamic bayesian network,DBN)是一种基于时序表达数据构建基因调控网络的重要方法。然而目前的DBN方法因计算时间太长,结构不稳定,准确度低,对有效性有很大影响。根据动态贝叶斯网络的度量可分解性质,将动态贝叶斯网络分为初始网络与转移网络分别进行结构寻优,在寻优时将基于静态贝叶斯网络的最大权重生成树算法与贪婪搜索算法相结合,移植入动态贝叶斯网络中,建立基因调控网络模型。提出了一种从时序数据中构建基因调控网络的方法,克服了贝叶斯网络不能描述循环调控的缺陷,也从规模上简化了网络构建问题。通过与相关实验文献的对照,验证了提出方法的有效性,网络学习时间明显缩短,网络结构更加稳定。
Dynamic Bayesian network (DBN) is an important approach for predicting the gene regulatory networks from time course expression data. However, three problems greatly reduce the effectiveness of current DBN methods, including long computational time, instable structures, and low accuracy. According to the property of decomposability of DBN, we divided DBN into initial network and transferring network, and combined Maximum Weight Spanning Tree algorithm (MWST) and Greedy Search(GS) algorithm both based on BN, then transplanted the mixed algorithm into DBN, to build the gene regulatory network model. We presented a gene regulatory network - building approach based on temporal expression data. Unlike previous DBN methods, our approach can described cycle - regulation among genes, and reduced the computational time in modeling the network. We verify the effectiveness of our approach by consulting the corresponding experiment articles, and reduce the computational time in learning network, finally get more stable structures.
出处
《生物医学工程研究》
2008年第3期145-149,共5页
Journal Of Biomedical Engineering Research
基金
国家自然科学基金资助项目(60471003)
关键词
时序表达数据
动态贝叶斯网络
度量可分解
最大权重生成树算法
贪婪搜索算法
基因调控网络
Temporal expression data
Dynamic Bayesian network
Measurement dividable
Maximum weight spanning tree algorithm
Greedy search algorithm
Gene regulatory network