HPC Lab - Software - BiBench

Table Of Contents

Previous topic

Welcome to BiBench’s documentation!

Next topic

Quick start

This Page


Instructions for installing BiBench and its dependencies. BiBench is written for Python 2. It has been tested with both Python 2.6 and 2.7.

Installation Procedure

Automatic Installation with pip

pip and virtualenv are currently the best way to install and manage Python packages. This method is also recommended because pip should automatically install BiBench’s dependencies. (Note: One of BiBench’s dependencies is rpy2, which requires that R be installed. R must have been built as a library. i.e., configured with --enable-R-shlib, and libR.so must be available in LD_LIBRARY_PATH. For more information, see the rpy FAQ).

If you want pip to install BiBench without installing dependencies, use pip install --no-deps in the following steps.

If it is not installed, install pip. If necessary, you can use pip to install virtualenv and virtualenvwrapper:

pip install virtualenv virtualenvwrapper

To use virtualenvwrapper you must source virtualenvwrapper.sh, so that its commands are available in Bash. One option is to add the following to your .bashrc:

source `which virtualenvwrapper.sh`

To create a new virtual environment called bibench_env and install BiBench and its dependencies to it:

mkvirtualenv --no-site-packages bibench_env
pip install http://bmi.osu.edu/hpc/software/BiBench-0.2.tar.gz

Pip can also install from a local package:

wget http://bmi.osu.edu/hpc/software/BiBench-0.2.tar.gz
pip install BiBench-0.2.tar.gz

BiBench is now installed and ready to go.

The BiBench package is now installed into its own virtual environment, bibench_env, which is an isolated Python environment. So before using BiBench you must switch to that environment:

workon bibench_env

When you are done working in that environment, simply run the deactivate command:


Manual installation

It is possible to manually install BiBench. In this case, all of BiBench’s Python dependencies must be installed seperately.

BiBench is packaged using distutils. To install BiBench manually, simply download it, unpack it, and run the distutils setup script (depending on your Python setup, you may need to be root for this step):

wget http://bmi.osu.edu/hpc/software/BiBench-0.2.tar.gz
tar xzf BiBench-0.2.tar.gz
cd BiBench-0.2
python setup.py install

Building the documentation

This documentation may be compiled into a number of formats, using sphinx. To generate html:

make html

For a list of all possible targets, run:

make help


The BiBench package is now installed, but most of its functionality will not be available until extra dependencies are available.

In particular, BiBench does not provide implementations for biclustering algorithms; they must be installed seperately.

Python Dependencies

If you install BiBench using pip, you should not need to install these packages manually; pip should automatically handle dependencies.

If you chose not to install BiBench’s dependencies before, you can install them with pip:

pip install numpy rpy2 decorator

To install optional dependencies simply run:

pip install nose sphinx

R Dependencies

Much of BiBench’s functionality depends on R. Some algorithms that BiBench supports are available as packages for R; BiBench also relies on R for some visualization methods and other functionality. To communicate with R, the Python package rpy2 must be installed.

rpy2 requires that R be compiled as a shared library using the configure option --enable-R-shlib. Also libR.so must be available in LD_LIBRARY_PATH. For more information, see the rpy FAQ.

Assuming that R is installed, entering the following commands in R should install all of BiBench’s R dependencies:

install.packages(c('biclust', 'isa2', 'MASS'))

The the rest of this section provides links to those dependencies.

The following may be installed with the R command install.packages():

The rest of the packages require Bioconductor. Here are the installation directions for Bioconductor and its packages.

For downloading GDS data:

For Gene Ontology enrichment analysis:

For missing data imputation:

Other Algorithm Dependencies

These algorithms are not available for R or Python. They must be manually built and installed, and their binaries must be available on the PATH, in order for the appropriate module in bibench.algorithms to work.