Reconstruction Of Gene Regulatory Network Based On Two Level Ensemble Learning

With the development of high-throughput technologies,a large number of gene expression data have been generated and accumulated.Therefore,the research direc-tion of biology has shifted to the gene level.The reconstruction of gene regulation network from gene expression data can help understand the relationship between gene expression and the basic function of organisms,which is of great significance for analyzing the complex life phenomena of organisms.Most existing methods re-construct gene networks with correlation coefficient,which cannot distinguish the direct and indirect interactions between genes.To deal with this problem,we ap-ply Gaussian graphical models to reconstruct gene networks.However,Gaussian graphical models are based on the assumption that the data follows a Gaussian distribution.When this assumption is violated,the estimated network may include some false positive edges.In addition,when the sample data is poisoned,the network accuracy will be affected.In order to solve the problem that estimation accuracy may be affected by the noise of data,this study puts forward a two layer ensemble learning method of constructing gene regulation network based on the idea of ensemble learning.The first layer is to resample sample data sets and generate multiple data sets and the second layer uses different robust estimation methods for each sub-sample data set to integrate the estimation results.The main feature of this method is to reduce the influence of noise and improve the stability of the result through multiple integra-tion experiments.Based on the algorithm of solving Gaussian graphical model,the simulation experiment demonstrates the improvement of accuracy in estimating the gene network when our robust method is applied.At the same time,the method is more stable for the noise data.