Reaction/Metabolite Standardization and Congruency Across Databases and Genome-Scale Metabolic Models

Databases and genome-scale models account for a rapidly expanding list of biotransformations occurring in nature. Available information can vary significantly but generally includes reactant and product designation, stoichiometric coefficients, organism assignment, compartment localization and associated genes. Unfortunately, both databases and metabolic models suffer by a number of shortcomings that complicates their use in metabolic model reconstruction, gap filling, and strain optimization efforts. Many inadvertently include reactions that may be elementally and/or charge unbalanced and the presence of multiple synonyms for the same metabolite and/or reactions among databases and models further complicates the sharing of information from multiple sources. In this talk, we describe how we used naming synonyms and molecular geometry comparisons to generate a reconciled database that encompasses BRENDA, KEGG, MetaCyc, React and over 30 genome-scale metabolic models. All reactions in the database are charge and elementally balanced, allowing for their direct use in simulations. This dataset as of today (Sept. 2010) consists of 34,984 distinct metabolites participating in 67,944 reactions.

We will discuss challenges associated with the handling of stereochemistry, non-specific side-chains, polymerization, and the identification of lumped reactions. A workflow that relies on a number of diverse methods is put forth to address these challenges. New methods use acoustic, textual, structural, reaction and network information in order to reconcile global data into a single unified knowledgebase. This knowledgebase allows for the direct elucidation of common/unique functionalities across genome-scale metabolic models and databases. We highlight the utility of the newly developed knowledgebase in gap-filling existing models, reconstructing new ones, elucidating shared pathways among organisms, and prospecting new pathway routes for bio-production. We also discuss the database technologies employed to efficiently store and query the developed knowledgebase.