Reconstruction of Gene Networks Using Phenomic Algorithms

Abstract

The reconstruction of gene networks has become an important activity in Systems Biology. The potential for better methods of drug discovery and disease diagnosis hinges upon our understanding of the interaction networks between the genes. Evolutionary methods are proving to be successful in such problems and a number of such methods have been proposed. However, all these methods are based on processing of genotypic information. We present evolutionary algorithms for reconstructing gene networks from expression data using phenotypic interactions, thereby avoiding the need for an explicit objective function. Specifically, we implement the Phenomic algorithm and validate it for the reconstruction of gene networks. We also extend the basic phenomic algorithm to perform multiobjective optimization for gene network reconstruction. We apply both these algorithms to the yeast sporulation dataset and show that the algorithms can effectively identify gene networks. Both the algorithms are validated for stability and accuracy in the reconstruction of gene networks.