Now beta version of rdkit is available from anaconda! So I would like to try it. However I would like to test without contaminating current my environment. So I tried new version of rdkit with Docker.
Fortunately rdkit can be installed via conda, so I made Dockerfie based on miniconda3. Following dockerfile used continuumio/miniconda3. By using the image I could install required packages via conda command. ;)
FROM continuumio/miniconda3 MAINTAINER iwatobipen RUN pip install --upgrade pip RUN conda install -c rdkit/label/beta rdkit -y RUN conda install -c conda-forge seaborn pandas scikit-learn -y RUN conda install -c conda-forge networkx jupyter jupyterlab -y RUN pip install py2cytoscape WORKDIR /workdir EXPOSE 8888 ENTRYPOINT ["jupyter-lab", "--ip=0.0.0.0", "--port=8888", "--no-browser", "--allow-root", "--NotebookApp.token=''"] CMD ["--notebook-dir=/workdir"]
Next, build image and run!
Type following command.
$ sudo docker build -t rdkit-beta . $ sudo docker run -it -p 8888:8888 --rm --name rdkit-beta --mount type=bind,src=`pwd`,dst=/workdir rdkit-beta
Then jupyter-lab will launch and I can access notebook the following URL. ‘localhost:8888’
And I tried to use new function named rdScaffoldNetwork. It is the implementation of scaffold tree. If reader would like to know what is the scaffold tree, please check the original article.
It is very useful for analyze compounds scaffolds. Following code is only just used it with simple case. So it is not practical example for real drug discovery project. I would like to apply the approach against real project near the feature.
I pushed dockerfile and notebook to my repo.
Any comments, suggestion will be greatly appreciated. And thank developer of RDKit!!! ;)