基因调控网络的边预测  

Link Prediction for the Gene Regulatory Network

作　　者：黎妍 张晓飞[2] 易鸣[3] 刘妍岩[1] 

机构地区：[1]武汉大学数学与统计学学院,武汉430072 [2]华中师范大学数学与统计学学院,武汉430079 [3]中国科学院武汉物理与数学研究所,武汉430071

出　　处：《数学物理学报（A辑）》2015年第5期1018-1024,共7页Acta Mathematica Scientia

基　　金：国家自然科学基金(11275259;91330113)资助

摘　　要：为构建基因调控网络,提出了一个基于基因表达水平和网络反传递的算法.该算法用网络反传递思想来分析由传统相关性计算方法产生的间接效果,并考虑了调控网络的稀疏性,在模型算法中加入了控制网络稀疏性的l_1范数惩罚项.在大肠杆菌实验数据上测试该算法,这种方法提高了相关性分析对调控网络中边的预测能力,皮尔逊相关系数提高了6.42%,斯皮尔曼相关系数提高了5.92%,互信息提高了9.35%.总的来说,这个模型为修饰大量系统的相关性数据提供一种新思路,可以应用到网络边的预测和推断生物网络的控制动力学中.In order to build gene regulatory network,we proposed an algorithm based on gene expression levels and network anti-delivery.The algorithm used the network anti-delivery idea to analyze indirect effect produced from the traditional correlated calculation methods,and added an norm penalty term for controlling network sparsity after considering the sparsity of regulatory networks.We use the algorithm on the E.coli experimental data.This method improves the edge predictive ability of the correlation analysis on the regulatory network,Pearson correlation coefficient increased by 6.42%,Spearman correlation coefficient increased by 5.92%,mutual information improves 9.35%.Overall,this model provides a new idea on modifying a large number of system-related data,and can be applied to network edge prediction and the control dynamics of biological network inference.

关 键 词：基因调控网络 直接相关 间接相关 皮尔逊相关系数 斯皮尔曼等级相关系数 互信息. 

分 类 号：O29[理学—应用数学]