基于有序条件互信息和有限父结点构建基因调控网络

发布时间：2018-02-09 03:48

本文关键词： 基因调控网络贝叶斯网络模型有序条件互信息有限父结点因果定向　出处：《生物化学与生物物理进展》2017年05期 　论文类型：期刊论文

【摘要】：基因调控网络重建是功能基因组研究的基础,有助于理解基因间的调控机理,探索复杂的生命系统及其本质.针对传统贝叶斯方法计算复杂度高、仅能构建小规模基因调控网络,而信息论方法假阳性边较多、且不能推测基因因果定向问题.本文基于有序条件互信息和有限父结点,提出一种快速构建基因调控网络的OCMIPN算法.OCMIPN方法首先采用有序条件互信息构建基因调控相关网络;然后根据基因调控网络拓扑先验知识,限制每个基因结点的父结点数量,利用贝叶斯方法推断出基因调控网络结构,有效降低算法的时间计算复杂度.人工合成网络及真实生物分子网络上仿真实验结果表明:OCMIPN方法不仅能构建出高精度的基因调控网络,且时间计算复杂度较低,其性能优于LASSO、ARACNE、Scan BMA和LBN等现有流行算法.
[Abstract]:Gene regulation network reconstruction is the basis of functional genome research, which is helpful to understand the mechanism of gene regulation, explore the complex life system and its essence. Only small-scale gene regulatory networks can be constructed, while the information theory method has more false positive edges and can not speculate on gene causality. This paper is based on ordered conditional mutual information and finite parent nodes. In this paper, an OCMIPN algorithm for rapid construction of gene control network. OCMIPN is proposed, which firstly uses ordered conditional mutual information to construct gene regulation related network, and then limits the number of parent nodes of each gene node according to the prior knowledge of the topology of gene control network. Using Bayesian method to deduce the network structure of gene regulation, The simulation results on artificial synthetic network and real biomolecular network show that the proposed method can not only construct a high-precision gene control network, but also have a lower computational complexity. Its performance is better than the current popular algorithms such as ARACNEScan BMA and LBN.
【作者单位】：西北工业大学自动化学院信息融合教育部重点实验室;宝鸡文理学院物理与光电技术学院;
【基金】：国家自然科学基金资助项目(91430111,61473232,61170134)~~
【分类号】：Q811.4;TP18

【相似文献】