Installing rdkit

post by ThetaPhiPsi on May 18, 2021

Just for the record, I am working on a cloud GPU instance and want to install rdkit for molecular computations.

But this isn’t so easy as one can see here: Installation — The RDKit 2023.09.2 documentation

I can’t install it via

pip install rdkit

and there’s no conda available (which I could install afaik know), but even the Ubuntu packages are not available on my instance server:

$ sudo apt-get install python-rdkit librdkit1 rdkit-data

Any ideas?

post by sdbrown on Jun 15, 2021

It’s not too hard to build RDKit from source pulled from github, and then add it to a venv. The RDKit build instructions are pretty decent, and only requires specifying the “alternate” build/install locations e.g. cmake -D PYTHON_LIBRARY=/usr/lib/python3.6/config/libpython3.6.a -D PYTHON_INCLUDE_DIR=/usr/include/python3.6/ -D PYTHON_EXECUTABLE=/usr/bin/python3 ..

I had to go through a couple iterations, but in the end I had it working in an otherwise pip-controlled environment without headache.

post by ThetaPhiPsi on Jun 16, 2021

Thank you, I’ll try that next time.

post by gonzcolm on Jun 23, 2021

Why don’t you just install conda/miniconda in your lambda workstation and then rdkit in a conda environment?

post by ThetaPhiPsi on Jun 30, 2021

I needed a quick solution back then and outsourced all rdkit parts to my laptop. This was possible as there were only pre- and post-processing steps using rdkit.

Thanks for the suggestion though.