Inferring Gene Regulatory Network with Recurrent Neural Network and Extended Artificial Bee Colony Algorithm

Abstract

Gene regulatory network is the network of genes interacting with each other performing as functional circuitry inside a cell. Many cellular processes are controlled by this network as they govern the expression levels of genes or gene product. High performance computational techniques are needed to analyze these data as it is heavily affected by noise. There are a number of algorithms available in the literature which use recurrent neural network for model building together with differential evolution, particle swarm optimization or genetic algorithm for searching the regulatory network. The problem with these methods is that they may trap in the local minimum. In this paper, we present an algorithm using recurrent neural network as model and an extended artificial bee colony algorithm for searching regulatory network that can avoid local minimum. A comprehensive analysis on both artificial and real data shows the effectiveness of the proposed approach. Furthermore we have also varied the network dimension and the noise level present in gene expression profiles. The reconstruction method has successfully predicted the underlying network topology while maintaining high accuracy. The proposed approach has also been applied to the real expression data of SOS DNA repair system in Escherichia coli and successfully predicted important regulations.