Installation¶
IDPConformerGenerator uses only Python-based APIs for which we expect it to run native on any system Python can run, as long as the third-party installation requirements are met.
Please note that SPyCi-PDB and
X-EISDv2 can be installed on top of the idpconfgen
Python environment. It is actually recommended since they both have IDPConfGen as a dependency.
We tested IDPConfGen on Ubuntu 18.04 LTS and 20.04 LTS as well as on WSL2.0 and the Graham cluster, an HPC resource of the Digital Research Alliance of Canada (DRAC).
Follow the steps below to install IDPConformerGenerator (idpconfgen or
IDPConfGen for short) on your local machine:
From source¶
Clone from the official repository:
git clone https://github.com/julie-forman-kay-lab/IDPConformerGenerator
And navigate to the new IDPConformerGenerator folder:
cd IDPConformerGenerator
Note
The requirements.yml describe the Python dependencies of
IDPConformerGenerator. If you are skilled managing python environments you
can go on your own. Otherwise, you can calmly follow our install
instructions.
At the end of the installation process, you will have a miniconda3
directory inside the IDPConformerGenerator main directory where the whole
installation is placed. If your ever want to delete IDPCG from your computer,
simply delete the IDPConformerGenerator directory.
To install IDPConfGen, run the following three commands. Wait until one finishes before running the second one:
./install_miniconda3.sh
source activate.sh
./install_deps.sh
Once this finishes, idpconfgen is ready to be used. Go to the usage and continue from there.
Remember to active the idpconfgen environment every time you open a
new terminal window. For that, navigate to the IDPConformerGenerator
repository folder and source the activate.sh file:
cd path/to/my/IDPConformerGenerator
source activate.sh
Update¶
To update to the latest version, open a new terminal window, and navigate to the
IDPConformerGenerator source folder. Remove the miniconda3 environment:
rm -rf miniconda3
Update the source to the latest version:
git pull
Reinstall the project and it’s dependencies. Run the following commands, one after the other, wait for them to finish before running the next one:
./install_miniconda3.sh
source activate.sh
./install_deps.sh
Your installation will become up to date with the latest developments. If you
had installed MC-SCE, Int2Cart, SPyCi-PDB, or X-EISDv2, you need to reinstall
them again in the idpconfgen environment.
From source on the Graham Cluster (DRAC)¶
Log-in and make sure you are in your user home directory:
cd
Load the required python packages and modules on Graham’s servers:
module load scipy-stack dssp boost
Create and activate a virtualenv as DRAC recommends anaconda3
not be installed in the home folder:
virtualenv --no-download idpcgenv
source idpcgenv/bin/activate
For the first time installation, install dependencies manually using pip.
Please note that the --no-index searches through DRAC’s available packages.
If they’re not available, it will install from the web:
pip install --no-index --upgrade pip
pip install numba --no-index
pip install pybind11 --no-index
pip install tox
We are ready to clone from source and installation from here will be similar to local:
git clone https://github.com/julie-forman-kay-lab/IDPConformerGenerator
cd IDPConformerGenerator
Make sure you’re in the idpcgenv virtual environment before
installing. Install with:
python setup.py develop --no-deps
When you login again to your cluster account remember to reactive the
idpconfgen environment before using idpconfgen:
cd
source idpcgenv/bin/activate
Installing third-party software¶
Some functionalities of idpconfgen require third-party software. These
are not mandatory to install unless you want to use such operations.
DSSP¶
IDPConfGen uses DSSP to calculate secondary structures. However, you only need DSSP if you are generated the database from scratch. If you use a prepared database JSON file you don’t need to install DSSP.
Please note we are only compatible with DSSP versions 2 and 3. If you have
installed DSSP version 4 (check by using the command mkdssp --version) please
refer to this issue
for a proper re-installation after removing DSSP version 4.
Install MC-SCE¶
IDPConformerGenerator can integrate MC-SCE to generate sidechains on top of the backbone conformers it generates, on the fly. For that you need to install MC-SCE on top of the idpconfgen Python environment. First, install IDPConfGen as described above. Next, follow these steps:
# ensure you are in the parent IDPConformerGenerator GitHub folder
# clone and enter the MC-SCE GitHub repository
git clone https://github.com/THGLab/MCSCE
cd MCSCE
# Make sure you're in the idpconfgen environment then
# install the additional dependencies using pip
pip install tensorflow
pip install tqdm
pip install pathos
# Install MC-SCE on top of IDPConformerGenerator
python setup.py develop --no-deps
# Navigate back to the IDPConformerGenerator GitHub folder and install
# `idpconfgen` again if needed
cd ../IDPConformerGenerator
python setup.py develop --no-deps
Now, if you choose the flag -scm mcsce in idpconfgen build command,
IDPConfGen will use MC-SCE to build sidechains as backbone conformers are
generated. You will see idpconfgen build -h has a specific group of
parameters dedicated to MC-SCE, you can explore those as well.
For installation on a cluster via virtualenv, dependencies need to be manually installed as the following for MC-SCE:
# ensure you're in the idpcgenv and the IDPConformerGenerator GitHub folder
git clone https://github.com/THGLab/MCSCE
# MC-SCE also requires numba and tox but that's already handled in previous steps
pip install tensorflow --no-index
pip install keras --no-index
pip install tqdm --no-index
pip install pathos --no-index
# cd into the MCSCE GitHub folder and install MC-SCE
cd MCSCE
python setup.py develop --no-deps
# cd back into the IDPConformerGenerator GitHub folder and install idpconfgen on top of MC-SCE
cd ..
python setup.py develop --no-deps
Install Int2Cart¶
IDPConformerGenerator can use Int2Cart on the fly to optimize bond geometries
of the backbones calculated. For this feature, you must have a CUDA compatible
GPU as well as install Int2Cart on top of the idpconfgen Python environment.
First, install IDPConfGen as described above. Next, follow these steps. Please
note that these steps are the same if you have installed idpconfgen through virtualenv:
# ensure you are in the IDPConformerGenerator GitHub folder
# Install a pre-requisite of Int2Cart: sidechainnet
git clone https://github.com/THGLab/sidechainnet
cd sidechainnet
pip install -e .
cd ..
# Install Int2Cart
git clone https://github.com/THGLab/int2cart
cd int2cart
pip install -e .
pip install pyyaml
cd ..
# you should be back in the IDPConformerGenerator GitHub folder
Running Int2Cart on the Graham cluster requires GPU allocations and module load cuda.
Otherwise, installation is the same within the idpconfgen virtualenv.
Troubleshooting Int2Cart installation¶
If IDPConfGen is still giving you an error that Int2Cart is not installed, please test this import
in the idpconfgen environment:
python
>>> from modelling.models.builder import BackboneBuilder
If you receieve this error: ImportError: TensorBoard logging requires TensorBoard version 1.15 or above,
do the following:
pip install tensorboard==1.15.0
CheSPI¶
To use CSSS via the idpconfgen csssconv command you need CheSPI. Please
refer to https://github.com/protein-nmr/CheSPI to install CheSPI.
δ2D¶
The use δ2D via the idpconfgen csssconv command you need δ2D.
Please refer to https://github.com/carlocamilloni/d2D.
Installing back-calculators and reweighting protocols¶
Both SPyCi-PDB and X-EISDv2 have been developed in-house with considerations
for protein structural ensembles in mind. We recommend to install both of
these packages on-top of the idpconfgen environment for streamlined usage.
Install SPyCi-PDB¶
Clone the SPyCi-PDB repository to the parent directory of where IDPConformerGenerator was cloned:
git clone https://github.com/julie-forman-kay-lab/SPyCi-PDB
Activate the idpconfgen environment and install the missing dependencies:
pip install pandas
pip install natsort
Move into the SPyCi-PDB directory and install on top of IDPConfGen:
cd SPyCi-PDB
python setup.py develop --no-deps
Install X-EISDv2¶
Clone the X-EISDv2 repository to the parent directory of where IDPConformerGenerator was cloned:
git clone https://github.com/THGLab/X-EISDv2
Activate the idpconfgen environment and install the missing dependencies.
You can skip this step if you’ve already installed SPyCi-PDB:
pip install pandas
pip install natsort
Move into the X-EISDv2 directory and install on top of IDPConfGen:
cd X-EISDv2
python setup.py develop --no-deps