METATOOL 5.0 for GNU octave and MATLAB


This version of Metatool consists partly of script files which are compatible with octave and Matlab and partly of shared libraries that are specific for the operating system and math program. In principle the script files provide all necessary functionality, but usage of the shared libraries makes elementary mode calculation far more efficient. The purpose of this distribution is to make the operations performed by Metatool more easily understandable and to allow the users to quickly adapt and extend the scripts according to their needs. Caveat: Due to currently sparse documentation and too few commentaries in the scripts this may not be so easy after all.



Extract the common files (common-files.tar.gz) and platform/program specific files (see table below) into one directory. Do not mix the specific files for the different math programs in one directory.

GNU octave 2.1.72 (api-v13)
MATLAB 6 or higher
Linux x86 octave-specific-files.tar.gz matlab-linux-specific-files.tar.gz
Windows n/a matlab-windows-specific-files.tar.gz
Mac OS X

The Linux libraries require and as well as some other standard libraries (type e.g. 'ldd elmo.oct' for the octave installation to check if all dependencies are resolved).

Metatool for FluxAnalyzer

If you want to use the Metatool option for calculation of elementary modes in FluxAnalyzer, you need to download the appropriate MATLAB-specific files for your operating system. Then extract the elmo.* shared library from the archive and copy it into the main installation directory of FluxAnalyzer. In addition, you have to put the kernel.m function from the common files into this directory.


Using standard Metatool input files

The standard Metatool input format is described here. Start your math program and cd into the directory where you have placed the files (or tell the program to include the directory in its search path). Then execute:

ex= parse('example.dat');

This will read the input file and store its content into several fields of the data structure 'ex' (the field names are: 'st', 'irrev_react', 'ext', 'ext_met', 'int_met', 'react_name'; cf. table below). If you get an error when executing this command, look in the known problems section below. The 'parse' command alone is useful when you want to manipulate the input data before the calculations. You can now call

ex= metatool(ex);

to perform the calculations or if you want both steps in one go you can directly call

ex= metatool('example.dat');

This will perform most of the operations that the stand-alone Metatool does and stores the results in the variable 'ex'. If you run metatool with a second argument as in

ex= metatool('example.dat', 'example.out');

then it will produce an ouput file similar to the one of the stand-alone version. There are two drawbacks with this kind of output: First, it is not possible to load this output later again. Second, for large systems the process of writing this file is very time-consuming. Therefore it is often better to use the built-in functions of the math program to save and load your data.
The variable 'ex' (or whatever you choose to call it) is a structure that contains several fields, the most important of which are:

st stoichiometric matrix (rows correspond to internal metabolites, columns to reactions)
irrev_react row vector which contains 0 for a reversible and 1 for an irreversible reaction
kn kernel (nullspace) of the stoichiometric matrix
sub subset matrix (rows correspond to the subsets, columns to the reactions in st)
rd reduced system
irrev_rd reversibility of the subsets in the reduced system
rd_ems elementary modes of the reduced system (rows correspond to the subsets, columns are elementary modes)
irrev_ems row vector which contains 0 for a reversible and 1 for an irreversible elementary mode
ext same structure as st, but rows correspond to external metabolites
int_met names of the internal metabolites
ext_met names of the external metabolites
react_name names of the reactions

When you call metatool with a second argument so that an output file is produced, there will also be a field calles 'ems' in the return variable. This field contains the elementary modes of the full system. When you run metatool without producing an output file, the elementary modes for the full system are not produced. The reason for this is to save memory since the the number of elementary modes in large/complex networks can explode. If you want to expand the elementary modes to the original system do this:

ex.ems= ex.sub' * ex.rd_ems;

Of course, this will produce one large matrix. But you can also choose which modes to expand by using the standard indexing mechanisms (cf. math program documentation), e.g.

ems_part= ex.sub' * ex.rd_ems(:, 1:10);

will only expand the first ten elementary modes. To examine the numerical quality of the result you can look at something likes this:

max(max(abs(ex.rd * ex.rd_ems)))

Ideally, the result should be zero or a very small number. If all your stoichiometric coefficients were integers and the result is not 0, something went wrong.

Using a stoichiometric matrix directly

In case you don't have your network as a metatool input file but want to calculate the elementary modes for a stoichiometric matrix directly, this is also possible. You simply have to set up a variable with the two fields 'st' and 'irrev_react' in the same way as described in the table above. Here is an example how to calculate the elementary modes of a network with two metabolites and four reactions: [1 1 -1 0; 0 0 1 -1];
net.irrev_react= [1 0 1 1];
net= metatool(net);

After calculation, net.rd_ems contains the elementary modes of the reduced system. Since no reaction and metabolite names were given, no metatool output can be produced.


Known problems

Determination of math program fails

Because octave and Matlab are not 100% compatible, there are some parts in the scripts where it is necessary to determine which math program is used. This is done by setting the global variable 'MATH_PROGRAM' to '0' for octave and '1' for Matlab. However, there are slight differences in the various octave and Matlab versions concerning the declaration and definition of global variables. As a result, the determination of the math program can fail leading to different consecutive errors. If you encounter an error when running Metatool, set the global variable 'MATH_PROGRAM' manually and retry. In order to do this, you first have to declare:


and then assign this variable '0' if you are using octave and '1' if you are using Matlab.

Parsing of Metatool input file fails

The Metatool input files are in ASCII format and it is essential that the linefeed encoding used by them corresponds to conventions of the operating system you are working on. To convert between UNIX and DOS formats use the unix2dos and dos2unix utilities. In addition, the last line in the input file must end with a linefeed.

Selected publications

S. Klamt, J. Gagneur, A. von Kamp: New algorithmic approaches for computing elementary modes in large biochemical reaction networks, IEE Systems Biology, in press.

T. Pfeiffer, I. Sánchez-Valdenebro, J. C. Nuño, F. Montero and S. Schuster:METATOOL: For Studying Metabolic Networks. Bioinformatics 15, 1999, 251-257 (describes important concepts, but not the current implementation).

S. Schuster, D. Fell and T. Dandekar: A General Definition of Metabolic Pathways Useful for Systematic Organization and Analysis of Complex Metabolic Networks. Nature Biotechnology 18 (3), 2000, 326-332. PubMed (general introduction to elementary modes analysis; the algorithm described in this paper is not used for the current implementation).

R. Urbanczik, C. Wagner:An improved algorithm for stoichiometric network analysis: theory and applications. Bioinformatics  21, 2005, 1203-1210 (basis for the null space algorithm used in the current implementation).


If you have any questions or comments and especially when you encounter errors please tell me.

Last updated: 19.7.2006