Flux Assignment (with) LAD Convex Objectives (and) Normalization. LAD, or Least Absolute Deviations, is the 1-norm equivalent of least squares.
Currently, two algorithms are implemented. The first algorithm estimates enzyme complex abundance, which is fed in to the second algorithm that also takes a genome-scale metabolic model as input. The output of the second algorithm are estimated fluxes for the environment that the expression and model were taken from.
====
FALCON should work on all major platforms (Windows, Linux, OS X). At this time it is required that the user is familiar with the UNIX command-line and has some experience installing UNIX software. It is possible that it could work on other platforms as well, but some of the MATLAB code would need to be ported to GNU Octave. In principle should be straightforward.
There are several essential prerequisites:
-
MATLAB: currently required for running the flux-fitting algorithm and running most of the analyses. This is the only non-free requirement.
-
COBRA Toolbox: A MATLAB package developed for analysis of genome-scale metabolic models. We assume the COBRA standards and use many of its functions in our code. See the official page for documenation and the protocol. We have used a fork of the project available at Ronan Fleming's repository.
-
Boehm GC: This should be available on most systems, though in Cygwin it will have to be explicitly installed using the setup program. The headers may need to be installed separately, which is necessary if one wishes to compile the minimum disjunction program.
-
libgcc This is the GCC runtime library, and should be installed on most systems by default. Sometimes the appropriate version may need to be installed if you are using an executable of minDisj compiled by someone else (if this is the case it is best you compile your own copy of minDisj).
-
C compiler: We have only tested recent versions of GCC. This is needed for building the implementation of the first algorithm in the pipeline (minimum-disjunction). For Windows, we suggest using GCC in Cygwin, as this is also currently needed for an optional but desirable prerequisite: ATS.
-
ATS: The first algorithm is written in ATS2, and while it should be possible to compile a version of it from the C sources we have generated, if one runs in to problems the easiest thing to do will be to try to compile from ATS. Nonetheless, please report any bugs encountered in compiling from C sources in the GitHub issue tracker. ATS2 can be installed by following directions on the ATS site and the ATS2 wiki. Users interested in altering or extending minDisj or GPR-related work file will need to install the ATS language, which has a style similar to StandardML.
After having acquired MATLAB and having installed the COBRA Toolbox, all that is needed is to install the minimum-disjunction program. It may be worth trying out a provided binary, which are available for some platforms. These are not guaranteed to be up to date or to work on all distributions or versiosn of systems. If it does not work, please try building it instead, using either the provided C sources or the ATS sources. Should building fail, please submit a report to the GitHub issue tracker.
Copy or create a link to the file with the name minDisj, and make sure it is in your PATH.
Once minDisj is installed, consider adding the following to your startup.m file (in UNIX, this should be in ~/Documents/MATLAB/), or run it before each use of FALCON within MATLAB:
addpath(genpath('MY_FALCON_DIRECTORY'));
This will be especially useful if you plan on running analyses similar to those found in the publication; otherwise, the top-level directory should suffice.
Use this method if the provided binaries do not work for you or there are none matching your operating system, and if you aren't interested in experimenting with GPR-related aspects of the algorithm.
Download the latest ATS includes (click on the latest version, e.g. ats2-postiats-x.y.z, then download the file (ATS2-Postiats-include-x.y.z.tgz). Extract the archive, e.g.
tar zxvf ATS2-Postiats-include-x.y.z.tgz
Set the environmental variable PATSHOME to the new directory:
cd ATS2-Postiats-include-x.y.z
export PATSHOME=`pwd`
Go in to the GPR subdirectory and build falcon_cmdline
cd <MY_FALCON_DIRECTORY>/GPR/MYPORTDIR
make -C .. MYPORTDIR
make PATSCC=gcc MYCCRULE=PORTABLE -f ../Makefile
And as stated above, copy or link to this file in your path as the file minDisj.
After installing ATS2 (see above), download the latest source from this repository.
Go in to the GPR subdirectory and run
make falcon_cmdline
And as stated above, copy or link to this file in your path as the file 'minDisj'.
For GPR related analyses, you can use minDisj on the command-line or, if you wish to use minDisj from MATLAB, the wrapper computeMinDisj.m will be of help. Also, see computeMinDisj.m for a description of the required expression file format.
For other data and information related to FALCON and and associated pubications, please see:
http://openwetware.org/wiki/Barker:Notebook/FALCON
In particular, data for testing the minimum-disjunction program can be found there.