Using MetRxn for flux elucidation and model reconstruction

Akhil Kumar2* (azk172@psu.edu), Saratram Gopalakrishnan1* (sxg375@psu.edu) and Costas

D. Maranas1

1Department of Chemical Engineering, Pennsylvania State University, University Park, PA; 2

The Huck Institutes of the Life Sciences, Pennsylvania State University, University Park, PA;
http://maranas.che.psu.edu/
MetRxn (http://www.metrxn.che.psu.edu/) is a searchable electronic resource for published metabolic models and databases from a wide variety of organisms. The MetRxn project aims to organize and disseminate standardized metabolite and reaction information to improve metabolic modeling by accurately describing reaction stoichiometry, directionality, atom mapping from reactants to products, and gene to protein to reaction relations. Standardization algorithms automatically curate information to remove incompatibilities in content representation, fix stoichiometric errors such as elemental or charge imbalances and resolve incomplete atomistic details. The standardization procedure follows a workflow that starts by matching metabolite entries using lexicographic and phonetic techniques, and structure comparison using atomistic details. The reactions are first charge and mass balanced and subsequently atom/bond mapping resolution algorithms are applied. For each reaction, metabolite stoichiometry, atom transition and metabolite compartment information is stored. The reaction and metabolite information is downloadable in SBML 3.0 and in a tabular format. The current MetRxn update includes recently published metabolic data for a total of 112 metabolic models and 8 metabolic databases. The number of distinct reactions that have been mapped is greater than 20,000 and MetRxn contains tools that allow users to download atom mapping data for each reaction.
As part of our ongoing effort we have enhanced the MetRxn knowledgebase with additional information such as reaction atom transition information and reaction standard free energies. In accordance with our data integration goal, we have integrated the ncbi taxonomy database, uniprot gene idâ??s and ncbi gene idâ??s within MetRxn. We developed a customized algorithm for quickly generating unique molecular graphs and detecting symmetries for all metabolites in the database. This is used to create atom transition information between reactants and products for all reactions contained in MetRxn. This information is leveraged for the construction of genome- scale size mapping models to support metabolic flux elucidation using C13 labeled substrates through metabolic flux analysis (MFA).
Models used for MFA typically include only central metabolism and simplified reactions for amino acids synthesis. We decided to explore how including additional reactions may affect the flux inference process using MFA. To this end we first appended onto a popular core metabolic model amino acid synthesis and degradation reactions from the iAF1260 model and we de- lumped a handful of central metabolism reactions. This modest modification led to significant changes in the confidence level for glycolysis and TCA cycle reactions and non-zero flux
through arginine degradation. Computational challenges and results obtained upon expansion to a core genome-scale model for E. coli will be described.

Supported by funding from the U.S. Department of Energy to Dr. Costas D. Maranas grant DE- FG02-05ER25684.