基于加性噪声模型的基因调控网络构建算法

Additive noise model based gene regulatory network construction algorithm

摘要:

为在统计推断方法通过相关性来筛选基因对时,能够体现调控关系的因果性,受因果定向算法能够有效定向调控关系的启发,将加性噪声模型与因果定向算法相结合,用基于加性噪声的定向算法度量因果关系的程度,提出了一种基因调控网络构建的算法. 该算法首先将加性噪声模型的因果定向算法扩展为一个特征选择算法,并通过建立调控因子集合与每个基因间的加性噪声模型来选择基因的调控因子. 在DREAM5的3个数据集上的实验结果显示,结果比其他算法有明显提升,该算法可有效构建基因调控网络.

Abstract:

In order to represent causal relationship when relevance measure is used in statistic inference methods to filter gene pair, inspired by the research that casual-effect orientation algorithm can identify direction of causal-effect variables effectively,we propose an additive noise model based on the gene regulatory network construction algorithm by using additive noise model orientation algorithm to measure degree of causal relationship. The algorithm extends additive noise model based orientation algorithm to a feature selective algorithm, and builds ANM model of transcription factors set and each gene to select transcription factors of gene. In the experiments of three datasets DREAM5, the method has clear improvement in comparison with other algorithms, and could be used as an efficient algorithm to build gene regulatory networks.