The 13CFLUX2 Modeling Plug-In extends a network model by certain properties specific for 13CFLUX2. 13CFLUX2 is a high performance simulator framework for isotope-labeling-based metabolic flux analysis developed at the public research institute IBG-1, Forschungszentrum Jülich.
Table of contents
- 1.2. Using the Plug-In
- 1.2.1. Model Setup
- 1.2.2. 13CFLUX2 Model Properties
- 1.2.3. Editing Configurations
- 126.96.36.199. Basic Properties
- 188.8.131.52. Constraints
- 184.108.40.206. Network Input
- 220.127.116.11. Measurement
- 18.104.22.168. Variables
- 1.2.4. Validity Check
- 1.2.5. Advanced Configuration
- 1.3. Special Application Cases
- 1.4. Further Reading
13CFLUX2 is a simulator framework for the quantification of metabolic fluxes from experimental measurement data with high-performance computational methods. The framework consists of about twenty different command-line applications for simulation, exploration, parameter fitting, statistical analysis and other purposes.
Simulation with 13CFLUX2 bases on models provided in the simulator specific FluxML format, an XML-based model format that stores network stoichiometry, atom mappings, constraints, start conditions, configurations of the network, measurement specification and data, simulation settings and others.
Basically, the entire modeling process for 13CFLUX2 can take place in Omix. The entire 13CFLUX2 model consists of different model parts that are realized by a number of different plug-ins:
- network stoichiometry - the stoichiometry of the network model is a built-in feature of Omix.
- atom mappings - the Atomic Layer Plug-In provides modeling features to define carbon atom mappings for each reaction of the network. Optionally the Chemical Structure Plug-In allows to equip the network with chemical structures that make the atom mappings unique and well interpretable. If a metabolite provides a structural formula its InChi code is exported as annotation of the pool in FluxML.
- stoichiometric constraints - the Network Constraints Plug-In equips the Omix document with the capability to define constraints for net and exchange fluxes as well as pool sizes. (The global bounds parameter available in the Network Constraints Plug-In is skipped in FluxML.)
- model configuration - the 13CFLUX2 Modeling Plug-In provides extended modeling features for the 13CFLUX2 model prepared in Omix. This includes:
- isotope enrichment of the network's input substrates
- measurement data e.g. MS or NMR measured isotope enrichments of the network's intermediates
- model variables and properties
- simulation settings
Subsequently, the different modeling features of the 13CFLUX2 Modeling Plug-In are introduced in detail.
After activation the list of network properties contains another entry "13CFLUX2 Model Properties".
If your network is compartmentalized, the same metabolite in two different compartments (including the non-compartment area) result in different pools in the exported 13CFLUX2 model (FluxML). Note, that cofactor edges, effector edges and additional graphical elements in the diagram are fully ignored by the FluxML exporter.
Finally, the completed network stoichiometry must have input pools and effluxes in order to be accepted by 13CFLUX2. An input pool is a metabolite that is only connected on the educt side of unidirectional reactions, i.e. no connection goes into the metabolite (cf. image a). An efflux is a unidirectional reaction without products, i.e. no connection goes out of the reaction (b).
Activate the Network Constraints Plug-In on the document in order to define further stoichiometric constraints for the network model.
In order to define atom mappings activate the Atomic Layer Plug-In and follow the instructions of the plug-in documentation. For 13CFLUX2 models it is absolutely recommended to use chemical structures as basis for the atom mappings. By using chemical structural, the atom mappings are based on well-defined atom positions in the reactant molecules. Furthermore, if chemical structures are available, they are used when defining input labeling and labeling measurements. This warrants consistency between the atom mappings and the labeling model because it is based on the same unique and reproducible atom identification algorithm.
The KEGG and BioCyc plug-ins allow to load the metabolic pathways of entire organisms from corresponding databases. The imported model also contains atom mappings and chemical structures.
In 13CFLUX2, it is not important for simulation of isotope labeling experiment what chemical element is actually used for isotope labeling. But important is, that the weight difference between the main isotope and the labeling isotope is 1:
- 12C, natural abundance 98.9%
13C for labeling
- 16O, natural abundance 99.8%
17O for labeling
- 14N, natural abundance 99.6%
15N for labeling
The Atomic Layer Plug-In supports the definition of atom mappings for all element types simultaneously. The "Simulated Element" restricts the considered atom mappings to the selected atom type. Hence, only the atom mappings are exported to FluxML that correspond to the "Simulated Element".
Each model should provide a model configuration named "default". However, arbitrary different configurations can be specified which must be explicitely referenced to be used when starting the 13CFLUX2 tools (see simulator documentation). By default, 13CFLUX2 uses the "default" configuration.
By selecting the configuration in the list and clicking the "Edit" button you can start editing the model configuration. You can also add a new empty configuration, duplicate the selected one and import model configurations from FluxML files.
- Basic Properties
- Network Input
- Include measurement model (checked by default)
- No measurement data
- Experimental design
Network Constraints Plug-In). How to use the editor to define stoichiometric constraints is described here.
If the "Network Input" section does not contain any input pool entries 13CFLUX2 automatically initializes the isotopomer distribution of each input source according to the natural abundance of 12C (98.945%) and 13C (1.055%) isotopes (even when the simulated element is not carbon).
You can add input labeling entries to the model by pressing the "Add" button. In the following dialog, you have to select which input metabolite you wish to specify (a). The selection dialog only lists actual input pools of the network because selecting an intracellular pool as input will be rejected by 132CFLUX2 as an invalid model.
After selecting an input pool a second dialog window appears where you can define the fractional abundance of a certain labeling pattern in the input source (b). Here, you have to specify which atoms of the input substrate are labeled. Therefore, click on the corresponding atom in the scene. If a structural formula is available for the selected input pool, you define the labeling pattern based on the displayed chemical structure which warrants clear atom identification.
After confirming the dialog, the new input pool and labeling pattern is listed. Alternative, you can choose "Default Labeling Pattern". When no labeling pattern is listed, 13CFLUX2 assumes natural carbon isotope abundance for the input pool (see above).
You can add multiple labeling patterns for each input substrate ("Add Labeling Pattern" button) leading to a labeling mixture. In this case, specify the ratio of the individual labeling patterns in the mixture by editing the value column.
For each labeling pattern you can optionally specify the positional isotopic purity of labeled carbon atom positions as value between 0.0 and 1.0. If an atomic purity is entered, all unlabeled carbon atom positions are assumed to be naturally labeled (i.e. 98.945% 12C, 1.055% 13C). If no atomic purity is given, they are taken 100% 12C and 0% 13C.
Instead of specifying the labeling patterns as isotopomer (labeled and unlabeled positions) you can also use the cumomer notation (states: labeled, unlabeled and either/or). Therefore, select "Cumomer" in the "Type" combobox on top.
The "Measurement" page contains different tabs for labeling and flux measurements. In modeling non-stationary isotope labeling experiments, also poolsize measurements can be defined.
However, a more convenient way to define a labeling measurement group is by using the graphical features of the modeling plug-in. Therefore, click on the "Add measurement group" button in the cell which opens the following wizard dialog:
Here, you can select the measured pool from a list of all metabolites in the network and choose which measurement method has been applied:
- 1H NMR
After specifying a measurement group in the first column, insert measured value and standard deviation in the second and third column. The fourth column represents the "scale" attribute of the measurement group. This attribute controls whether the measurement values are automatically scaled (value "auto") or not (value "1").
First, specify the measured molecule fragment. Therefore, click on the individual atoms in order to select which belongs to the fragment (green) and which does not (red). Then, select the mass isotopomers that have been measured by clicking on the rectangular buttons (M0..Mn) on bottom.
After finishing, the table is filled with the measurement group expression and additional rows for each mass isotopomer (M0..Mn).
In order to define the MS/MS group, you must first specify the measured molecule fragment in two fragmentation steps. Therefore, click on the individual atoms of the mother ion (M=mass of mother ion) and daughter ion (m=mass of daughter ion) in order to define the fragmentation. Green atoms belong to the fragment, red one do not. Then, select the mass isotopomers that have been measured by clicking on the rectangular buttons (M0..Mp × m0..mq → M(0..p,0..q)) on bottom.
After finishing, the table is filled with the measurement group expression and additional rows for each mass isotopomer (M(0..p,0..q)).
More importantly it is possible to discriminate labeled from unlabeled atom positions in the neughborhood of hydrogens, and thus, this 1H-NMR allows the description of positional isotopic labeling enrichment. A 1H-NMR measurement group specification is essentially a list of labeling fractions at specified positions. It is assumed that all listed atom positions have at least one bond to a hydrogen atom.
Define a 1H-NMR measurement group by clicking on the individual atoms in order to select which position was measured (green) and which was not (red).
After finishing, the table is filled with the measurement group expression and additional rows for each labeling position (P1..Pn).
Therefore, select "Use Generic Labeling Pattern" and mark each atom of the metabolite to be measured (black), not measured (white) or "either/or" (shaded). After finishing, the content of the table cell remains editable and you can edit the generic formula. A generic measurement specification consists of a semicolon-separated list of formulas. Each of these formulas describes exactly one measurement value. The formulas contain isotopomer notations and support various operators ("+", "-", "*", "/", "yˆx") including some built-in functions (sqrt, abs, min, max, exp, log, log2, log10). Syntactical valid examples:
- E4P#x1x1 + E4P#1xxx
- E4P#0101; E4P#1000
- 2 * E4P#0x11 * log(E4P#x1x0)
Finish editing the generic measurement formulas by pressing the ENTER key or clicking on a different table cell. After finishing, the table is filled with additional rows for each independent formula in the generic group (separated by ";").
- CSV spreadsheets - available by default
- Excel spreadsheets - requires the Excel plug-in
- FluxML - requires the FluxML plug-in
- FWD - requires the FWDSIM plug-in
When importing measurement data from FWD file, you must specify the standard deviation. This is done in the dialog shown in the next picture. Here, you can specify a constant value as standard deviation for all imported values. Alternatively, you can specify a mathematical expression how the standard deviation shall be computed from the imported value x.
Furthermore, it is possible to copy-paste data from Excel tables or from other model configurations to the measurement tables.
On the "Variables" page, you can specify the variables and specify their initial values. The page provides a table for net and exchange fluxes. In non-stationary mode, you can also define free poolsizes.
Add a variable by clicking on a cell in the first column and select an entry from the list of reactions. Thereafter, you can specify an initial value in the second column.
The selection of the system variables must be valid, i.e. no variable may be selected that stoichiometrically depends on other variables. If the selected set of variables is incomplete, the simulator chooses additional variables with initial value 0. The 13CFLUX2 utility sscanner can be used to automatically determine an optimal selection of free fluxes with optimal initial values. After running sscanner, import the variables from the provided FluxML file in order to carry the variables over into the model.
Free Flux Plug-In to specify free fluxes within the network diagram. Every time you select a reaction as free, the Free Flux Plug-In recomputes the dependencies in the stoichiometry. By this, only those reactions can be selected that do not have any dependency.
You can import the free fluxes selection from the Free Flux Plug-In by selecting the "Import from Free Flux Plug-In" button on the toolbar.
- CSV spreadsheets - available by default
- Excel spreadsheets - requires the Excel plug-in
- FluxML - requires the FluxML plug-in
- FWD - requires the FWDSIM plug-in
- HDF5 - requires the HDF5 plug-in
It is also possible to copy-paste data from Excel tables or from other model configurations to the variables tables.
The consistency check includes:
- Correct specification of input pools and network effluxes (see section Model Setup)
- Specification of constraints
- Specification of measurements and measurement data
- Specification of variables
The attribute "Simulation Type" controls the behavior of network reduction:
- "Auto" (the default) selects automatic network reduction. This results in choosing the optimal simulation model based on either a Cumomer or EMU network for the specified measurement model. This should be used by default to obtain the fastest possible simulation.
- "Full" deactivates automatic network reduction. As a consequence a simulation of the full / unreduced Cumomer or EMU network is enforced - although this might not be necessary. Using this setting, simulations are usually much slower or even impossible for large networks comprising metabolites with many labeling positions. Use with care.
- Checking "Explicit" results in evaluation of the measurement model, explicitly listing the measurement specifications that are to be simulated. This option does neither require nor allow actual measurement values. Therefore, specify "No measurement data" in the "Measurement" box on bottom of the "Basic Properties" page. In this mode, all measurement values are assumed to be zero and all standard deviations are assumed to have the value one.
Furthermore, you can you can influence the applied simulation method. The default value is "Auto", which selects the optimal simulation method for the specified measurement model. Other possible choices are "Cumomer" (for all NMR and MS/MS measurements) and "EMU" (for MS measurements). A non-automatic selection of the simulation method makes sense combination with simulation type "Full".
In combination with a specification of a mixture of different available labeling substrates the program edscanner samples the possibly high dimensional mixing simplex at evenly spaced points. At each sampling point a linearized statistical analysis is performed and a chosen single-numbered measure (design criterion) is calculated providing information on the high dimensional ellipsoid given by the fluxes' covariance matrix is computed. The coordinates of the sampling points including the information criterion are written to a HDF5 file. A further analysis of this data may include the visualization of mixing triangles.
Based on a 13CFLUX2 model with an optimized flux distribution the program edopt can be used to generate an optimal design for a subsequent isotope labeling experiment. For this purpose, edopt performs a linearized statistical analysis and determines the mixture of substrates which minimizes the criterion value for the flux distribution’s covariance ellipsoid, i.e. by using the A-criterion it allows the most accurate flux determination having the smallest sum of standard deviations (see technical documentation of 13CFLUX2 for futher information). In contrast to edscanner, edopt is not limited to choosing the equidistant points in mixture triangle, but tries to optimize the substrate mixture directly to obtain a minimal A-criterion.
In order to perform experimental design, make sure, that the network stoichiometry only contains one input pool. Specify a number of input labeling states for this input pool. Optimally, you specify three different labelings in order to use the mixing tiangle visualization.
Each of the labeling states should be represented by one item in the left list. Each state should have one labeling pattern entry in the right list. Do not insert the labeling mixture to be evaluated in experimental design under the same input state. Each labeling pattern has furthermore an optional "Cost" attribute. The actual costs of labeling patterns are considered in experimental design while finding an optimal labeling mixture.
In case of a multi-value measurement group (e.g. multiple MS mass isotopomers) you can specify the error model in the row of the measurement group specification or in each value row. In the first case, the error model is used for all measurement values in the measurement group. In the second case, each measurement value requires an error model. That is, it is possible to specify different error models for individual measurement values.
In case the error model is completely omitted no extrapolation of the standard error is performed and the original provided standard error is used when computing the covariance matrix required for experimental design. The error model may be set to "1" for disabling the scaling of the covariance matrix completely.
The attribute "Weight" takes a floating-point value between 0.0 and 1.0 (inclusive) and is used to adjust the influence of a free flux on the elements of the covariance matrix. Setting "Weight" to 0.0 elimates the influence of a free flux. Any value between 0.0 and 1.0 results in a scaling of the corresponding row and column of the covariance matrix. The default value of the attribute is 1.0.
The non-stationary model parts are experimental and subject to be changed or removed.
- Wiechert W. (2001) 13C metabolic flux analysis. Metab Eng 3(3):195-206
- Wiechert W, Möllney M, Isermann N, Wurzel M, de Graaf AA (1999) Bidirectional reaction steps in metabolic networks. Part III: Explicit solution and analysis of isotopomer labeling systems. Biotechnol Bioeng 66:69–85
- Wiechert W, Wurzel M (2001) Metabolic isotopomer labeling systems. Part I: Global dynamic behaviour. Math Biosci 169:173–205
- Weitzel, M., Wiechert, W., and Nöh, K. 2007. The topology of metabolic isotope labeling networks. BMC Bioinformatics, 8 (1), 315.
- Nöh, K., Weitzel, M., and Wiechert, W. 2008. From Isotope Labeling Patterns to Metabolic Flux Rates. In U. H. E. Hansmann, J. H. Meinke, S. Mohanty, W. Nadler, and O. Zimmermann (Eds.) From Computational Biophysics to Systems Biology (CBSB08), vol. 40 of NIC Series, (pp. 345–348).