Reconstructing Gene Regulatory Networks from Homozygous and Heterozygous Deletion Data Using Gaussian Noise Model

Abstract

A transcription network is a large and complex network consisting of interacting elements that, over time, affect each other's state. The dynamics of complex gene regulatory processes are difficult to understand using intuitive approaches alone. To overcome this problem, many research studies have been conducted for reconstructing gene regulatory networks (GRN). In this research study, we propose an algorithm for inferring the regulatory interactions from homozygous and heterozygous deletion data using a Gaussian model. Using simulated gene expression data on networks of known connectivity, we investigate the ability of the proposed algorithm to predict the presence of regulatory interactions between genes and the signed edges (activation or suppression). The algorithm is applied to network sizes of 10 genes and 50 genes for two E.coli subgroups and three S.cerevisiae/ Yeast subgroups. The predicted networks were evaluated on the basis of two scoring metrics, area under the ROC curve (AUROC) and area under the precision-recall curve (AUPR). The algorithm has reconstructed the networks with a reasonably low error rate. Our AUPR and AUROC values are consistently higher than the other method compared in this study. The Gaussian model distinguishes real signals from random fluctuations using an iterative method. The analysis of the experiment results reveals that our method can reconstruct networks and predict signed edges with a wide range of network types, connectivity, and noise levels with a reasonable error rate.