TeachOpenCADD: a teaching platform for computer-aided drug design (CADD) using open source packages and data
Dominique Sydow, Maximilian Driller, Andrea Morger and Andrea Volkamer
In silico Toxicology, Institute for Physiology, Universitätsmedizin Berlin, Virchowweg 6, 10117 Berlin
This work is licensed under the Attribution 4.0 International (CC BY 4.0). To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
(Back to Table of contents.)
Open source programming packages for cheminformatics and structural bioinformatics are powerful tools to build modular, reproducible, and easy-to-share pipelines for computer-aided drug design (CADD). While documentation for tools is available, only few freely accessible examples teach underlying concepts focused on CADD usage such as the TDT initiative [1], addressing especially users new to the field.
Here, we present a CADD teaching platform developed from students for students, using open source packages such as the python tools RDKit [2], PyPDB [3], and PyMol [4]. For each topic, an interactive Jupyter Notebook [5] was developed, holding detailed theoretical background of the underlying topic and well-commented python code, freely available on GitHub. Illustrated at the example of EGFR kinase, we discuss how to acquire data from the public compound database ChEMBL [6], how to filter compounds for drug-likeness, and how to identify and remove unwanted substructures. Furthermore, we introduce measures for compound similarity, which are subsequently used to cluster compounds, i.e., for the selection of a divers compound set. We also employ machine learning approaches in order to build models for predicting novel active compounds. Lastly, structures are fetched from the public protein database PDB [7], used to generate ligand-based ensemble 3D pharmacophores and to conduct binding site comparison for off-target prediction.
With this platform, we aim to introduce interested users to the ease and benefit of using open source cheminformatics tools. Topics will be continuously expanded and are open for contributions from the community. Beyond their teaching purpose, the notebooks can serve as starting point for users’ project-directed modifications and extensions.
Literature:
[1] S. Riniker et al., F1000Research, 2017, 6, 1136 [2] G. Landrum, RDKit [3] W. Gilpin, Bioinformatics, 2016, 32, 156-60 [4] The PyMOL Molecular Graphics System, Version 1.8, Schrödinger, LLC [5] T. Kluyver et al., IOS Press, 2016, 87-90. [6] A. Gaulton et al., Nucleic Acid Res., 2017, 40, D1100-7 [7] H. Berman et al., Nucleic Acid Res., 2000, 28, 235-42
(Back to Table of contents.)
Topics on TeachOpenCADD are demonstrated in form of so called talktorials (a mixure of theory and coding, ready to be presented also in form of a presentation) and currently include:
- Compound data acquisition: ChEMBL
- Molecular filtering: ADME and lead-likeness criteria
- Molecular filtering: unwanted substructures
- Ligand-based screening: compound similarity
- Compound clustering
- Maximum common substructures
- Ligand-based screening: machine learning
- Protein data acquisition: Protein Data Bank (PDB)
- Ligand-based pharmacophores
- Binding site similarity and off-target prediction
Please contact andrea.volkamer@charite.de if you have question or suggestions on existing or potential new talktorials.
(Back to Table of contents.)
Use the Anaconda software for a clean package version management.
Install Anaconda2 or Anaconda3. In theory, it should not matter which Anaconda version you install. However, we only tested it for Anaconda2, so we cannot garantee the same behaviour for Anaconda3.
https://docs.anaconda.com/anaconda/install/
Then, use the package management system conda to create an environment for our talktorials.
Create an environment, e.g. called cadd-py36, and activate this environment.
conda create -n cadd-py36 python=3.6 anaconda
conda activate cadd-py36Install required packages within this environment.
Note: In order to link your environment to your jupyter notebook, install nb_conda (see below). After starting your jupyter notebook from the terminal within your environment, select your environment via Kernel > Change kernel > Python [conda env:cadd-py36].
https://stackoverflow.com/questions/39604271/conda-environments-not-showing-up-in-jupyter-notebook https://stackoverflow.com/questions/37433363/link-conda-environment-with-jupyter-notebook
conda install jupyter # Is probably already installed
conda install nb_conda # Link conda environment with jupyter notebook
# If nb_conda does not work, try out ipykernel:
#conda install ipykernel # Is probably already installed
#python -m ipykernel install --user --name cadd-py36 # Enable the environment in jupyter notebook
conda install -c rdkit rdkit
conda install -c samoturk pymol
conda install -c conda-forge pmw # Necessary for PyMol terminal window to pop up
conda install -c conda-forge pandas
conda install scikit-learn
conda install -c anaconda seaborn
pip install biopandas # or conda install biopandas
# pip is probably installed by default in your environment
pip install chembl_webresource_client
pip install pypdbYou can download the talktorials to your computer using the package git:
git clone https://github.com/volkamerlab/TeachOpenCADD.gitWithin your conda environment cadd-py36 start Jupyter notebook...
jupyter notebook... and get started with the first talktorial. Enjoy!
Generally, the installation works the same under Linux and Windows without problems.
The only difference, we encountered, is the installation of PyMol:
- Download PyMol from http://www.lfd.uci.edu/~gohlke/pythonlibs/#pymol
- Put it where Anaconda2 is installed (e.g. C:\Anaconda2)
- Open a command prompt as administrator (Windows > accessories > command prompt, "right-click" > open as administrator)
- Go to the Anaconda directory by typing the correct path on the prompt:
cd C:/Anaconda2
set path=%path%;C:\Anaconda2
pip install pymol‑2.3.0a0‑cp36‑cp36m‑win_amd64.whl
pip install pymol_launcher‑2.1‑cp36‑cp36m‑win_amd64.whl
# or whatever pymol_XXXX.whl you have downloadedfrom rdkit.Chem.Draw import IPythonConsolethrowsImportError: No module named 'PIL': try to install pip install pillow in your respective environment (check out rdkit/rdkit#1179)Draw.MolsToGridImagethrows some Serial error: list your input data type as inDraw.MolsToGridImage(list(df.ROMol), useSVG=True)
The package manager pip under version 10.0.X works differently than before, so that the command pip install chembl_webresource_client suggested above throws the import error cannot import name main:
You can solve this problem with calling pip via python like this:
python3 -m pip install chembl_webresource_client
python3 -m pip install pypdbInstalling the chembl_webresource_client, you might run into the warning that urllib3 is incompartible. You can solve this via updating:
python3 -m pip install urllib3 --upgradeSolution is taken from: