Steps to Install SimSite3D¶

Setup a Virtual Environment for SimSite3D¶

Note

You may skip this step if you have a recent version of Python 2 (2.6+) and you want to install the SimSite3D Python dependencies either in your home directory or in the system (e.g. /usr/local).

Note

you only need to setup the tools virtualenv and virtualenvwrapper once per machine if you install it in a standard location. Of course, you may have multiple virtual environments, but don’t need to replicate the two tools used to work with virtual environments.

Choose a directory to store the virtual environments. This should be a directory that is group read/writable and is relatively low latency or local to the machine. On the PSA network mounts, this location is currently:

/soft/linux64/.virtualenvs

Install a recent version of Python¶

Go to python.org and get the latest 2.* version and install it. Currently the installed version is at:

/soft/linux64/Python-2.7.2

This version was installed as an alternate install and should probably only be used for virtualenvs.

Install a recent version of Boost¶

Go to boost.org and get the lastest stable Boost version and install it. Currently the installed version is at:

/soft/linux64/boost_1_48_0

Build the boost::python libary using latest version of Python:

./bootstrap.sh --with-libraries=python --with-python-root=/soft/linux64/Python-2.7.2 --prefix=/soft/linux64

Edit the tools/build/v2/site-config.jamfile (in the installed boost directory) so that we use this version of boost::python and the right python library:

project site-config ;

using gcc ;
# We are using an alternate location install of Python 2.7
using python : 2.7 : /soft/linux64/Python-2.7.2 ;
# We had to build boost using the alternate install of Python for ....
lib boost_python : : <file>/soft/linux64/boost_1_48_0/stage/lib/libboost_python.so ;

# These are dependencies for SimSite3D -- if someone felt the need they could
# only link them where needed.
lib blas : : <file>/usr/lib64/libblas.so.3 ;
lib lapack : : <file>/usr/lib64/liblapack.so.3 ;
lib popt : : <file>/usr/lib64/libpopt.so ;

Edit the SimSite3D/python/boost-build.jam to point to this version of boost build; currently this may be done as:

boost-build /soft/linux64/boost_1_48_0/tools/build/v2 ;

Install virtualenv¶

Next we want to install virtualenv Get virtualenv from the website:

cd /soft/linux64/src
wget http://pypi.python.org/packages/source/v/virtualenv/virtualenv-X.Y.Z.tar.gz

Unzip it in a src directory:

tar -xvzf virtualenv-X.Y.Z.tar.gz

Build and install it in a local directory:

cd /soft/linux64/src/virtualenv-X.Y.Z
/soft/linux64/Python-X.Y.Z/bin/pythonX.Y setup.py install --prefix=/soft/linux64/Python-X.Y.Z

Setup the virtualenv using python and virtualenv.py¶

Note

If you don’t want to use bash, you must follow this section as virtualenvwrapper is written in bash

Initialize a virtual environment using:

python virtualenv.py --no-site-packages --distribute <name or path of VE>

Edit the activate script to set the necessary environment variables¶

Set the environment variables in $VIRTUAL_ENV/bin/activate and/or $VIRTUAL_ENV/bin/activate.csh

Set the SIMSITE3D_INSTALL_DIR variable:

setenv SIMSITE3D_INSTALL_DIR=</path/to/SimSite3D>

Activate the virtual environment¶

Activate your desired virtual environment (virtualenv) by changing to that directory and sourcing the correct activate file (for your shell). If you are using csh/tcsh, you may use:

cd <path to my virtualenv>
source bin/activate.csh

Setup the virtualenv using virtualenvwrapper¶

Note

virtualenvwrapper is quite nice, but it isn’t necessary and it does require that you use the bash shell

Note

you only need to setup virtualenv and virtualenvwrapper once per machine if you install it in a standard location.

I chose to do it in this manner. Edit your environment and login files to set the following environment variables (using bash):

# the following are needed for virtualenv and virtualenvwrappers
export PYTHONPATH=$PYTHONPATH:/soft/linux64/Python-X.Y.Z/lib/pythonX.Y/site-packages
export WORKON_HOME=/soft/linux64/Python-X.Y.Z/.virtualenvs
source /soft/linux64/Python-X.Y.Z/bin/virtualenvwrapper.sh

Download, build, and install virtualenvwrapper:

cd /soft/linux64/src/
wget http://pypi.python.org/packages/source/v/virtualenvwrapper/virtualenvwrapper-2.11.tar.gz
tar -xvzf virtualenvwrapper-2.11.tgz
cd /soft/linux64/src/virtualenvwrapper-2.11
/soft/linux64/Python-X.Y.Z/bin/pythonX.Y setup.py install --prefix=/soft/linux64/Python-X.Y.Z

Initialize a virtual environment using:

mkvirtualenv --no-site-packages --distribute <name of new VE>

Activate the drugsite-prod virtualenv:

workon <name of new VE>

Installing SimSite3D¶

Before cloning the SimSite3D repository, we will need a recent version of mercurial. This can be done by installing mercurial in the virtualenv using pip:

pip install mercurial

Cloning the SimSite3D Mercurial Repository on BitBucket¶

Clone SimSite3D from BitBucket:

hg clone ssh://hg@bitbucket.org/jeff_vanvoorst/simsite3d <local dir name>

Change directory to the root directory of the repository:

cd <local dir name>

Install SimSite3D’s Requirements¶

SimSite3D’s code can be divided among three categories: pure C++ code, C++ code for boost::python extension modules, and Python code.

The core of SimSite3D is the pure C++ code

The core code has several dependencies that depend on what you seek to do * The molecular surfaces for binding sites require one of the following:

Generate surfaces yourself: * Generating surfaces using your tool of choice * Converting them to the format written by MSMS * Passing the surfaces to gen_points via command line flags

Having an MSMS binary at the path $SIMSITE3D_INSTALL_DIR/bin/linux_msms

Providing the location of an MSMS binary to gen_points via a command line flag

ArtSurf uses numerical linear algebra that should be as robust as reasonably possible * We chose to use LAPACK, and require:

A Fortran77 LAPACK library

A BLAS library that can be loaded by the LAPACK library

Boost::Python extension modules exist for some of the higher level C++ classes, and they require: * Boost headers * a recent version of Boost::Python * Python development tools (header files and libraries)

There are a number of utility functions, etc. that do not require high performance, but rather require less time coding; these tools are written in Python and will require: * numpy * matplotlib * sphinx * mercurial