摘要
为构建基因调控网络,提出了一个基于时序互信息学习动态贝叶斯网络结构的学习算法.在计算基因间的时序互信息时,该算法考虑了时间序列微阵列数据的时间特性,并利用协方差矩阵计算互信息,没有将基因表达数据离散化,与基因表达数据的连续性相符合.在酵母菌周期细胞的实验数据上测试该算法,灵敏度为66.7%;该算法构建的基因调控网络与KEGG数据库中的网络相比较,发现了Cdc28与Cdc20、Chk1与Rad9的调控关系,这些调控关系在相应的生物学实验中得到验证.
In order to construct gene regulation network, a learning algorithm was proposed based on dynamic Bayesian network with time series mutual information learning. In calculation of time series mutual information be- tween genes, the proposed algorithm takes into consideration the time property of time series microarray data and calculates mutual information with covariance matrix, so that the gene expression data are kept indiscrete, which is in accordance with their continuum. Test of the algorithm on experiment data of yeast cell cycle shows that the sensi- tivity of the algorithm was 66.7 %. In comparison between the networks in KEGG database and those constructed with the proposed algorithm, the regulation relationships between Cdc28 and Cdc20 and between Chkl and Rad9 were identified, which were verified by related biology experiments.
出处
《天津大学学报》
EI
CAS
CSCD
北大核心
2010年第7期655-660,共6页
Journal of Tianjin University(Science and Technology)
基金
天津市应用基础及前沿技术研究计划重点资助项目(07JCZDJC06700)
关键词
基因调控网络
动态贝叶斯网络
时序互信息
gene regulation network
dynamic Bayesian network
time series mutual information