Augmenting Scalability in ARACNe and CLR for Inferring Large Gene Regulatory Networks from Expression Profiles

Abstract

Inference of gene regulatory networks (GRNs) is an essential step towards understanding the complex interactions of an organism. In this regard, many methods have been proposed to infer gene regulatory networks over the years. However, most traditional methods do not apply to real networks because of their inability to infer large networks. The Algorithm for the Reconstruction of Accurate Cellular Networks (ARACNe) and the Context Likelihood of Relatedness (CLR) are popular methods for inferring GRNs. Their capability is restricted to a few hundred nodes despite their qualitative goodness. Therefore, we developed an effective parallel framework that augments the scalability of ARACNe and CLR to infer large networks in parallel without compromising their original predictive quality. In this paper, we describe the workflow and the results of the framework implementation for both methods. The framework is tested using synthetically generated DREAM Challenge networks. Results show that the new framework is highly scalable and effectively infers relatively large networks with no loss of true edges.