Personal tools
You are here: Home / Installing Biskit / Windows installation

Windows installation

At least the Biskit core library should also work on Windows. We just haven't really tested it on this platform. The interface to external programs may or may not work. Please report your experience!

1. Install core prerequisites

Most required packages have windows installers. Python 3.0 is currently not supported by Biskit. Please install:

  • Python 2.7
  • numpy (for python2.7)
  • Biopython (for remote fetching of PDBs)
  • SciPy (does not provide windows installer but several options are described)

It may be easier to install the Anaconda python bundle instead, which will give you all the above packages and many more in a friendly Windows installer. Example using Miniconda:

  1. download and install Miniconda

  2. run Anaconda shell (as Administrator if you have chosen a system-wide setup) and then execute:

    conda install numpy biopython scipy
    
  3. Test your Python installation:

    python
    >>> import numpy, scipy, Bio
    

Should give no errors.

Note: Beware of alternative Python installations that may exist on your system. Mixing up Python instances is a typical source of trouble.

2. Install Biskit

2.1 using pip

If you are using the anaconda or similar python distros, pip is already installed. Simply execute:

pip install biskit

2.2. manual setup, using latest git checkout

This has the advantage that you have the latest code and can quickly apply modifications.

  1. Clone Biskit into your user folder:

    git clone https://github.com/graik/biskit.git biskit
    
Alternatively, you could also fetch Biskit through the GitHub Desktop application.
  1. Ensure Python can find the Biskit package folder (the [capital] Biskit folder with the __init__.py file). You could either achieve this through a symbolic link or (more classically) using the PYTHONPATH variable. Let's assume your Biskit package is located in C:\Users\YourUser\Documents\biskit\Biskit. You can create a symbolic link into your Python's site-packages folder. If using Miniconda, this should look approximately like this:

    mklink /D C:\ProgramFiles\Miniconda2\Lib\site-packages\Biskit C:\Users\YourUser\Documents\biskit\Biskit
    

Or you put the location of the Biskit package folder into a PYTHONPATH environment variable: Right-click My Computer, then Properties (click) / Advanced (tab) / Environment Variables (Button). Then Add the variable PYTHONPATH with the value C:\Users\YourUser\Documents\biskit\Biskit as above.

2.2 via Windows installer

download and execute the windows installer (biskit*w32.exe) from the Biskit download area.

or... create your own windows installer (also works on unix) from the current master branch on github:

git clone https://github.com/graik/biskit.git biskit
cd biskit
python setup.py bdist_winst

The binary windows installer is then found in biskit/dist

Note:

The principal problem here is that you have little control which Python instance the package will end up in.

3. Install useful optional libraries

The above minimal installation lacks some functionality. Install also:

  • gnuplot -- for quick and dirty plotting (plot() and scatter()). The gnuplot binary must be in the search $PATH (see Notes below)
  • Plotutils + Biggles 1.6.5 -- for the advanced plotting functions that are built, for example, into PDBModel and Trajectory. This is a very useful extension but Biggles doesn't come with a windows installer and needs to be built from source using the conventional 'python setup.py install' command (after plotutils has been installed). Please report your experience!

4. Notes

Putting gnuplot into PATH:

Click your way to:

  • My Computer (right click) / Properties (click) / Advanced (tab) / Environment Variables (Button)
  • Edit entry $PATH and add ";C:pathtognuplotprogram"
Test:
open a "DOS" command line window. Type: gnuplot

This should give you the gnuplot Info screen:

G N U P L O T
Version 4.0 patchlevel 0
last modified Thu Apr 15 14:44:22 CEST 2004
...

Then open python and try again:

python
>>> from Biskit.gnuplot import*
>>> plot( range(10) )