Automatic identification and restoration of reaction gaps in the consensus reconstruction network for yeast metabolism

We present an automated procedure of identifying and filling of reaction gaps in genome-scale metabolic networks within the framework of flux balance analysis. This computational approach exploits the constraint-based optimisation techniques and graph traverse algorithm to identify the non-producible metabolites in the network and search for reactions to add into the model to restore the reachability of the metabolites or clusters of metabolites. This is a part of the iterative process of converting a genome-scale reconstruction into an executable computational model: representing the reactions in mathematical formats, validating and refining the mathematical model. We utilised this procedure for validation and refinement of YEASTNET2.0 (http://www.comp-sys-bio.org/yeastnet/), a recent version of the consensus reconstruction of yeast metabolism. The consensus reconstruction initially involves 1834 unique chemical reactions, 886 ORFs and 1418 metabolites located in 15 different compartments. 136 blocked metabolites of interests have been found non-producible from the network, and 117 (86.0%) of them can be restored by adding reactions from the reference models or putative transporter reactions. This one-step further computational effort over the initial manual curation towards a gapless network reconstruction model can systematically decrease the inconsistency of the model and potentially improve the accuracy of the model simulation. Furthermore, this approach can generate hypotheses (suggesting good candidate reactions) for manual verification or further experimental test.