chemreps is a Python package for the creation of molecular representations for the purpose of machine learning. The molecular representations included in this library are implemented/adapted from current literature. The aim of chemreps is to provide an easy to use library for making molecular representations that can be then used with machine learning packages such as Scikit-Learn and Tensorflow.

• Coulomb Matrix
• Bag of Bonds
• Bonds/Nonbonding, Angles, Torsions
• Just Bonds
• Morgan Fingerprints (RDKit Dependency)

The citations for the literature from which the representations are implemented/adapted from can be found in the source code for each representation.

Requests for new representations to be added can be made by raising an issue and labeling it as a feature request. Before requesting a new representation, please check under the Representation project in the Projects tab to see if that representation is included in the current work or progress.

The latest release version can be installed with:

pip install chemreps

The latest development version can be installed by:

git clone https://github.com/chemreps/chemreps
cd chemreps
pip install -e .

chemreps requires:

• Python (>=3.6)
• NumPy (>=1.12)
• cclib (>=1.5)
• QCElemental
• SciPy

• RDKit (for Morgan Fingerprints)

If you are interested in helping develop for this project, please check out Contributing to chemreps in the wiki for a guide on how to get started.

Tests can be run in the top-level directory with the command pytest -v --cov=chemreps tests/

If you need any help using chemreps, feel free to post in our Gitter.

• These are attempts at the recreation of molecular representations from literature and may not be implemented properly.
  • If we do not implement something properly, feel free to make an issue.
• This is solely a representation library and will not perform machine learning.

We now have a Zenodo release!