快速贝叶斯基因网络构建算法

摘要

Inferring the gene regulatory network is a major challenge in computational biology.During past decades, a lot of numerous computational approaches have been introduced for inferring the GRNs.Bayesian network methods cannot handle large-scale networks due to their high computational complexity, while information theory-based methods suffer from false positiveegative problems.To overcome the limitations, we present a novel algorithm.The algorithm first uses sequential conditional mutual information to construct initial networks.Then, the restriction of the maximum number parents for each gene is employed to generate gene regulatory network.The algorithm is tested on realistic biological networks and in silico networks of different sizes and topologies, and it outperforms other state-of-the-art methods.The results indicate that not only effectively reduces the computational cost due to much smaller sizes of local GRNs, but also considerably improves the precision of network inference.%基因网络构建是计算生物学一个很重要的研究领域,近年来涌现出大量推断基因网络构建的计算模型,各种模型方法都有自己的优缺点,如贝叶斯网络模型方法可以得出网络的最优结构,但是因其过高的计算时间复杂度只能应用于小规模网络;信息论的方法可以处理高维低样本数据,但构建出的基因网络中有过多的假阳性边.为了克服这些缺陷,提出了一种新的方法,该方法首先使用有序条件互信息构建基因调控的子网络,然后根据基因调控网络的拓扑先验知识,利用贝叶斯方法找出最优网络结构.该算法在计算机人工合成网络和真实生物分子网络上进行验证分析,其性能超过了现在流行的一些方法,试验结果表明,该方法不仅有较低的时间计算复杂度,而且也取得了较好的基因调控网络构建精度.