Personal tools
You are here: Home / Installing Biskit / Install prerequisites

Install prerequisites

Step-by-step instructions for installing all libraries and programs needed by the Biskit module itself.

These are the generic instructions for installing everything from source, which should work on any kind of Unix, including MacOS-X. There usually are short cuts depending on your flavor/distribution. Ubuntu Linux makes things particularly easy. See the Ubuntu-specific installation instructions. Even though we never cared for it, installation on Windows is actually also possible but requires some tricks -- contact us for help in this case.

Contents Detailed Instruction


Install required programs

Note, we assume that your computer has some standard development tools up and running. On Ubuntu, you can install the necessary programs like this:

sudo apt-get install gcc make autoconf libtool gawk
  • Python (version 2.5 required, 2.7 recommended)

    from rpm or source

    1. install from rpms or debian

      • python and (on many systems) python-devel (source / include files) (Ubuntu: python-dev)
    2. from source (works well on pretty much any system)

      • download from http://www.python.org/

      • in: python-2.x.x:

        ./configure
         make
         make install  (installs in /usr/local)
        

    Note:

    Redhat systems are notorious for shipping with old python versions that cannot easily be replaced. In this case, install the latest python from source into, e.g. /usr/local and then push the new python to the top of the search PATH, for example:

    ln -s /usr/local/bin/python2.5 ~/bin/python
    export PATH=~/bin:$PATH
    
  • Gnuplot

    from rpm or debian

    1. install the gnuplot package with your favorite package manager

    2. from source:

Install required libraries

General note:

Most python libraries are installed with a simple python setup.py install command and will end up in the site-packages folder of your local python installation. If you don't have sudo rights you can also install them into a different location (and set your PYTHONPATH appropriatly, for example:

python setup.py install --home ~/local

...will install your python library into ~/local/lib/python.

  • Numpy

    from rpm or source

    With Biskit 2.1 and higher we have switched to the new Numeric incarnation called numpy.

    1. from rpms (Ubuntu package: python-numpy)

    2. from source

  • NetCDF

    from rpm, deb or source

    Note:

    NetCDF is required by Scientific Python and will usually be installed automatically if you use Debian or RPM package managers.

    1. install from rpms (e.g. SuSE: package netcdf, Ubuntu: libnetcdf3)

    2. from source

      • download from http://www.unidata.ucar.edu/software/netcdf/ or from our package mirror (outdated)

      • in netcdf-3.6.0-p1/src:

        ./configure --prefix=/usr/local
        make test
        make install
        

        Note: On a recent install on an AMD 64bit system Scientific python failed to install if not netCDF was recompiled after setting:

        export CPPFLAGS="-fPIC -Df2cFortran"
        export FFLAGS="-fPIC"
        
  • Scientific python

    from rpm, deb or source

    1. install from rpms or debian package (e.g. Ubuntu: python-scientific)

    2. from source

  • Plotutils

    from rpm, deb or source

    1. install from rpms or debian package (Ubuntu: plotutils)

    2. from source

  • Biggles
    • check that x11 and libplot include+library files are installed (a separate x11-devel package in some distros, in Ubuntu you need libplot-dev):

      ~> ll /usr/lib/libXaw.so
      lrwxrwxrwx 1 root root 10 2007-05-16 14:36 /usr/lib/libXaw.so -> libXaw8.so
      
    • download and install biggles 1.6.6:

      wget http://downloads.sourceforge.net/biggles/python2-biggles-1.6.6.tar.gz
      tar xvf python2-biggles-1.6.6.tar.gz
      cd cd python2-biggles-1.6.6
      python setup.py build
      sudo python setup.py install
      
    • test:

      python
      >>> import biggles
      >>>
      

      Then paste the following code snippet into your python interpreter:

      import biggles, numpy
      x = numpy.arange(0,10,0.3)
      p = biggles.FramedPlot()
      p.add( biggles.Curve(x, numpy.cos(x)) )
      p.show()
      
    • Consult our biggles troubleshooting guide if you experience any problems!

  • BioPython

    from rpm, debian or source -- NOTE 3

    1. install from debian or rpms (e.g. Ubuntu: python-biopython)

    2. from source:

      1. mxTextTool

      2. BioPython

        • download from http://www.biopython.org/ (use latest version!)

        • in biopython-1.*:

          python setup.py build
          python setup.py test
          python setup.py install
          

Install PVM - Parallel Virtual Machine

Note: Feel free to skip the following steps if you don't care about parallelization.

  • PVM binaries

    from source or rpm

    • install pvm and pvm-devel from rpm or repositories (preferred!) (Ubuntu: pvm, pvm-dev)
    • from source: follow the PVM from source Howto!
  • setup PVM environment

    shell variables and links

    1. add to .zshenv (example, adapt to your shell):

      export PVM_ROOT=/usr/lib/pvm3  ## depends on system
      export PVM_RSH=/usr/bin/ssh
      export PVM_ARCH=`$PVM_ROOT/lib/pvmgetarch`
      

      (the variable $PVM_ARCH should match the name of a folder in $PVM_ROOT/lib/ describing the machine architecture -- LINUX or LINUX64 are typical values. In most cases, it is only needed for the compilation of pypvm and can be commented out afterwards. )

    2. In $PVM_ROOT/bin/$PVM_ARCH make links to applications that pvm needs -- biskit requires xterm and python, for example:

      cd $PVM_ROOT/bin/$PVM_ARCH
      ln -s `which xterm` .
      ln -s `which python` .
      
  • install pypvm

    python bindings for PVM

    1. download our repackaged pypvm:

      wget http://biskit.pasteur.fr/install/troubleshooting/pypvm_repackaged
      tar xvfz pypvm_repackaged
      
    2. in pypvm-0.95x-repacked:

      python setup.py build
      python setup.py install
      

      Note:

      this can turn into a bit of a headache if you run the latter command as sudo which then doesn't know about your $PVM_ROOT and $PVM_ARCH variables. Try the following (replace zsh and .zshrc by your default shell and shell configuration):

      sudo zsh
      source ~/.zshenv
      python setup.py install
      
  • password-less master -> slave access:

    ssh key generation

    1. on the master: If you don't yet have a .ssh/id_dsa.pub file, run:

      ssh-keygen -t dsa
      
    2. Check that there is no .ssh directory on the remote machine (otherwise skip the mkdir .ssh command in the next step).

    3. Copy public key from the master to the slave:

      cat ~/.ssh/id_dsa.pub | ssh user@machine "(mkdir .ssh;cat >> .ssh/authorized_keys2)"
      
    4. on the slave: run:

      chmod go-w . .ssh .ssh/authorized_keys2
      
    5. Check that passwordless login works without providing a password. -- NOTE 4

  • Test your PVM installation

    See PVM Howto!


Install Biskit library

  • get Biskit

    fetch a snapshot of the biskit package

    1. Fetch the most current version from SVN (anonymously):

      svn co https://biskit.svn.sourceforge.net/svnroot/biskit/trunk biskit
      

      Note:

      Check your firewall settings if you get "connection refused" or "timeout" errors at this step.

    2. Or developer access

      • If you are a biskit developer (contact us to become one!), use the following checkout command:

        svn co https://developername@biskit.svn.sourceforge.net/svnroot/biskit/trunk biskit
        

      (replace 'developername' with your sourceforge login)

    3. Or download last file release from SourceForge

      • Alternatively, you can also download the latest file release from sourceforge.

      • After downloading, unpack it with:

        untar xvfz biskit-2.0.0.tar.gz
        
  • Adapt PYTHONPATH

    the Biskit library folder needs to be added to the PYTHONPATH environment variable

    1. Edit your shell configuration file (example for .bashrc or .zshenv):

      export PYTHONPATH=$PYTHONPATH:~/biskit/
      

      (Assuming you checked out biskit into your home directory.)

Now you should be able to import the Biskit module in your python scripts.


Next steps


NOTE 3

  • The Ubuntu package python-biopython is usually one or two release numbers behind the actual version. If there are any issues with blast searches or other NCBI tools, try installing the current release from source.
  • [Resolved] If the biopython test hags at "test_SVDSuperimposer ..." you are suffering from the Numeric/eigenvalue problems. See Numeric Lapack+Atlas.

NOTE 4

Or download and use ssh-keyinstall to set up your system:
http://www.stearns.org/ssh-keyinstall/

Still having problems? Check for firewalls!


NOTE 5

[Resolved]:

  • Installing Scientific Python on 64bit based Linux systems might be tricky (tested with openSuse 11.1). In order to run this installation, you have to modify the setup.py of Scientific Python on line :

    line 48 : "lib/python%s.%s/site-packages/numpy/core/include" to : "lib64/python%s.%s/site-packages/numpy/core/include"

  • After this modification everything should go fine.