基于改进MS-EM算法的贝叶斯网络学习法构建基因调控网络  

Constructing Gene Regulation Network Based on Improved Bayesian Network Learning Methods of Model Selection Expectation Maximization Algorithm

作　　者：方跃胜[1] 姚宏亮[2] 

机构地区：[1]安徽水利水电职业技术学院,安徽合肥231603 [2]合肥工业大学计算机与信息学院,安徽合肥230009

出　　处：《佳木斯大学学报（自然科学版）》2014年第3期436-440,共5页Journal of Jiamusi University：Natural Science Edition

基　　金：国家自然科学基金资助项目(60705015);安徽省自然科学基金项目(KJ2013B095)

摘　　要：基因调控网络的重构是功能基因组中最具挑战性的课题之一.实验证明构建基因调控网络的最有前途的方法是贝叶斯网络.EM算法是一种有效的利用数据来学习贝叶斯网络的方法,能较好地处理构建基因调控网络中的数据缺失情况,但存在学习精度低、对初始参数值依赖的缺点.本文应用贝叶斯网络实现啤酒酵母细胞基因调控网络的构建,用改进的MS-EM算法进行学习,并实现实验结果的可视化.与现有文献比较,结果表明改进后的算法进一步降低了时间性能,提高了构建调控网络的精度.Modeling gene regulation network is one of the most challenging subjects of functional genome . According to experiments , the Bayesian Network is proved to be the most promising approaches .EM algorithm is an effective method in learning Bayesian network to use the data .Missing data in constructing gene regulation network are better dealt with EM algorithm .However , there is a low precision problem in learning and the result has strong dependence on the initial parameters .In this paper , Bayesian network was used to construct brewer＆#39;s yeast cell gene regulation network , and an improved EM algorithm was used to study .The visual process was a-chieved.Compared with the existing literature , the results show that the improved algorithm further reduced the time performance and improved the accuracy of constructing gene regulation network .

关 键 词：基因调控网络 贝叶斯网络 EM算法 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]