An automated approach to assigning subcellular localization in genome-scale metabolic network reconstructions

Abstract

Genome-scale models take a simplified view on metabolism by only considering the stoichiometry of the metabolic reactions. Despite this simplification, these models have proved useful in many different areas, such as in identification of metabolic engineering targets, for analyzing metabolite connectivity and pathway redundancy or for studying metabolic interactions between species. Due to the increasing popularity of this type of models, and the necessity to reconstruct models for less well-studied organisms, there is a need for tools to automate the reconstruction process. Recently there have been attempts to automate the network reconstruction based on protein sequence homology. However, these strategies are mainly aimed at prokaryotic systems where there is no subcellular compartmentalization of enzymes. Here we present a method for assigning subcellular localization to enzymatic reactions in an automated fashion. The algorithm aims at assigning localization in a manner that is consistent with signal peptide composition and physiochemical protein properties, while at the same time maintaining a well-connected and functional network. Non-enzymatic reactions, such as diffusion across membranes, are inferred based on connectivity. We believe that this technique can significantly speed up the otherwise very time consuming and laborious task of model reconstruction for eukaryotic organisms.