Installation¶
Platforms¶
Ensembler is developed and tested on Linux and Mac. Windows is not well supported.
Supported Hardware¶
The simulation refinement stages of Ensembler use OpenMM, which performs best on CUDA-enabled GPUs. However, OpenMM also has an optimized CPU platform, so use of GPUs is optional.
Install with Conda¶
conda is a package manager built for scientific Python. Unlike easy_install or pip, it handles binaries and binary dependencies, which are critical for most scientific workflows.
Conda can be obtained by installing Continuum Anaconda - an awesome free Python distribution for scientific computing. The standard installation contains many of Ensembler’s dependencies. Alternatively, for a more minimal Python set-up you can install miniconda, which contains only Conda and Python.
First get a license for Modeller from the Modeller website (registration required; free for academic non-profit use).
Save this in an environment variable as follows
$ export KEY_MODELLER=XXX
Then, to install Ensembler with Conda, use the following commands
$ conda config --add channels http://conda.anaconda.org/omnia
$ conda config --add channels http://conda.anaconda.org/salilab
$ conda install ensembler
Conda will automatically install all dependencies except for the optional dependencies Rosetta and MolProbity. These require licenses (free for academic non-profit use), and will have to be installed according to the instructions for those packages. Some limited installation instructions are included below, but these are not guaranteed to be up to date.
Install from Source¶
Clone the source code repository from github
$ git clone git://github.com/choderalab/ensembler.git
Then, in the directory containing the source code, you can install it with.
$ python setup.py install
Dependencies¶
To use Ensembler, the following libraries and software will need to be installed.
- Linux, Mac OS X or Windows operating system
- We develop mainly on 64-bit Linux and Mac machines. Windows is not well supported.
- Python >= 2.6
- The development package (
python-devorpython-develon most Linux distributions) is recommended.- Modeller
- Comparative modeling of protein structures.
- OpenMM
- Molecular simulation toolkit, with GPU-accelerated simulation platform.
- MDTraj
- Simulation trajectory analysis library.
- MSMBuilder
- Statistical models for biomolecular dynamics.
- PDBFixer
- PDB structure modeling
- BioPython
- Collection of Python tools for computational biology and bioinformatics.
- NumPy
- Numpy is the base package for numerical computing in Python.
- lxml
- For working with XML files.
- PyYAML
- For working with YAML files.
- docopt
- For building command-line interfaces.
- mock
- For testing in Python
Optional packages:
- MPI4Py
- Allows many Ensembler functions to be run in parallel using MPI.
- Rosetta
- Protein modeling suite. The
loopmodelfunction is optionally used by Ensembler to reconstruct missing loops in template structures.- subprocess32
- (If running Python 2.) Backport of the Python 3 subprocess module for Python 2. Used to run command-line programs such as Rosetta loopmodel. Includes timeout functionality which is particularly useful for the template loop reconstruction routine.
- Pandas
- Some functionality, including the
quickmodelandinspectfunctions, requires pandas.- MolProbity
- For model validation. The
package_modelsfunction can use this data to filter models by validation score.
Manually Installing the Dependencies¶
Linux¶
If you’re on ubuntu and have root, you can install most dependencies through your package manager (apt-get).
$ sudo apt-get install python-dev
Mac¶
If you’re on mac and want a package manager, you should be using homebrew and brews‘s Python (see this page for details). For example, numpy can be installed with brew as follows:
$ brew tap Homebrew/python
$ brew install python
$ brew install numpy
Then, you can install many of the remaining packages with pip.
$ pip install lxml
Windows¶
Chris Gohlke maintains windows binary distributions for an ever-growing set of Python extensions on his website. Download and install the the installers for setuptools, nose, numpy, scipy, numexpr, pandas and tables.
Testing Your Installation¶
Running the tests is a great way to verify that everything is working. The test suite uses nose and mock, which you can pick up via conda or pip if you don’t already have them.
$ conda install nose mock
To run the unit tests:
$ nosetests ensembler -a unit
Further tests are available which check interoperation of Ensembler with software dependencies such Rosetta loopmodel, or with external public databases such as UniProt, or are excluded from the unit tests due to being slow. To run them:
$ nosetests ensembler -a non_conda_dependencies -a network -a slow