Here we explore one of the major applications of steady-state metabolic modeling: the prediction of organism growth rates under varying perturbations. The two most common perturbations studied with metabolic models are variations in the nutrients available to the organism (e.g., changes in carbon source, nitrogen source, and oxygen availability), and the presence of gene knockouts. These two perturbations can be combined since the effects of gene knockouts can be modeled under different nutrient mixes.
Modeling Growth Under Varying Nutrient Conditions
Modeling of the preceding perturbations is accomplished by varying the inputs to the metabolic model. The nutrients available to the organism are specified in the nutrients section of the .fba file that defines a modeling run in the Pathway Tools MetaFlux module. Each nutrient is referenced using its unique identifier (frame id) in the Pathway/Genome Database (PGDB) from which the model is derived (see sample nutrients section for the EcoCyc metabolic model below). The unique identifier is followed by specification of the cellular location in which the nutrient appears, within square brackets. The cellular location is specified using our cellular component ontology (CCO). Although providing the cellular location of every nutrient may seem unnecessarily wordy, we have found it important to eliminate confusion as to where each nutrient appears. In the examples below, the nutrients are provided in the cytosol and in the periplasmic space. Upper and/or lower bounds on the allowable uptake rate of a nutrient by the model can optionally be provided.
GLC[CCO-PERI-BAC] :upper-bound 10
A key aspect of the flux-balance analysis methodology is to compute steady-state metabolic flux rates in the cell by calculating the maximal production of cellular biomass given the available nutrients and the supplied constraints on nutrient uptake rates. The cellular biomass components are specified by another section of the .fba file that lists the amino acids, nucleotides, fatty acids, and other products of cellular metabolism. MetaFlux translates this list of biomass metabolites into a large “biomass reaction” that lists biomass metabolites as its reactants, and “biomass” as its product. For example:
L-tryptophan + … + CTP + … + NAD + … --> biomass
The optimization procedure that calculates fluxes maximizes the rate of this biomass reaction. If the biomass reaction has a positive flux (meaning that every biomass metabolite is being produced by other reactions with a positive flux), then cell growth is predicted to occur. Furthermore, the cellular growth rate is the flux computed for the biomass reaction. Thus, a metabolic model predicts whether growth will occur under a specified set of nutrients, and it predicts the rate of growth.
Modeling Growth of Gene Knockouts
MetaFlux can model both gene knockouts and reaction knockouts under a given set of nutrients. The MetaFlux user interface allows the user to specify one or more genes to knock out in a model run, in which case MetaFlux removes from the model all reactions catalyzed by the products of those genes, if no active isozyme remains to catalyze that reaction. These gene/enzyme/reaction relationships are encoded within the PGDB. Because some gene products catalyze multiple reactions, and because of isozymes, MetaFlux also allows the user to specify the deletion of specific reactions. To enable high-throughput exploration of gene and reaction knockouts, the MetaFlux interface allows the user to request model runs involving all single-gene knockouts, double-gene knockouts, single-reaction knockouts, or double-reaction knockouts. Growth rates are calculated for every requested perturbation. For E. coli, calculating growth rates for every single gene knockout takes about two minutes on a 3 GHz workstation. The EcoCyc model predicts growth versus no-growth for gene knockouts with 95.2% accuracy.
Comparisons to Large-Scale Experimental Data
Because PGDBs can capture large amounts of experimental data on gene knockouts and on organism growth under different nutrient conditions (e.g., Biolog phenotype microarray results), we have written software that will compare the results of large-scale modeling predictions with large-scale experimental datasets within the PGDB to identify where prediction differs from experiment. Those disagreements provide opportunities for further refinement of a model, and for identifying noise in experimental datasets.
Learn More about MetaFlux
MetaFlux and the full Pathway Tools software are freely available from SRI for academic use. You can learn more about how to use MetaFlux by attending SRI’s metabolic modeling tutorials (the next tutorial is scheduled for March 18-19, 2015 in Menlo Park, CA), and by reading the Pathway Tools User’s Guide.