王广云 邹志康 季秀才 田燕 牛天慧.面向异步多时延基因调控网络建模的高阶动态贝叶斯网络模型及其结构
学习算法[J].现代生物医学进展英文版,2014,14(25):4958-4961. |
面向异步多时延基因调控网络建模的高阶动态贝叶斯网络模型及其结构
学习算法 |
High-order Dynamic Bayesian Network Model and It's Structure LearningAlgorithmfor Constructing Gene Regulatory Networks with AsynchronousMulti-time Delays |
|
DOI: |
中文关键词: 基因调控网络 异步多时延 高阶动态贝叶斯网络 学习算法 |
英文关键词: Gene regulatory network Synchronous multi-time delay High-order Dynamic Bayesian Network Learning algorithm |
基金项目:国家自然科学基金项目(81101177) |
|
Hits: 709 |
Download times: 821 |
中文摘要: |
目的:由基因芯片数据精确学习建模具有异步多时延表达调控关系的基因调控网络。方法:提出了一种高阶动态贝叶斯网
络模型,并给出了网络结构学习算法,该模型假定基因的调控过程为多阶马尔科夫过程,从而能够建模基因调控网络中的异步多
时延特性。结果:由酵母基因调控网络一个子网络人工生成了加入10%含噪声的表达数据用于调控网络结构学习。在75%的后验
概率下,本文提出的高阶动态贝叶斯网络模型能够正确建模实际网络中全部的异步多时延调控关系,而经典动态贝叶斯网络仅
能够正确建模实际网络中1/3的调控关系;ROC曲线对比表明在各个后验概率水平上高阶动态贝叶斯网络模型的效果均优于经
典动态贝叶斯网络。结论:本文提出的高阶动态贝叶斯网络模型能够精确学习建模具有异步多时延表达调控关系的基因调控网
络。 |
英文摘要: |
Objective:To precisely construct gene regulatory networks with synchronous multi-time delays from microarray gene
expression data.Methods:A high-order dynamic Bayesian network model and its structure learning algorithm were presented, the
network model assumed that the gene regulating process was high order Markov process, so it could model the synchronous multi-time
delays in gene regulation.Results:Artificial gene expression data with 10% noise were made from a sub-network of a yeast gene
regulatory network. With 75% posterior probability, the high-order Dynamic Bayesian Network model had correctly learned all the
regulatory synchronous multi-time delayed connections, while normal Dynamic Bayesian Network model had just learned 1/3 of all
correct regulatory connections. The receiver operator characteristics curves showed that with any posterior probability our model was
obviously much better than the normal Dynamic Bayesian Network model.Conclusion:The high-order Dynamic Bayesian Network can
precisely model tasynchronous multi-time delays in gene regulation, and more precise gene regulation networks can be learned from
microarray gene expression data by the high-order Dynamic Bayesian Network. |
View Full Text
View/Add Comment Download reader |
Close |