choderalab/charge-datasets

AM1-BCC ELF10 small molecule training datasets

Partial charge datasets for machine learning

This repository contains conformation-independent AM1-BCC ELF10 charge sets Generated using the OpenEye Toolkit 2019.4.2.

Author: John Chodera [email protected]

License

These datasets are licensed under the Open Data Commons Attribution License (ODC-By) v1.0 https://opendatacommons.org/licenses/by/1-0/index.html

Manifest

• environment.yml - conda environment
• datasets/ - datasets

Datasets

Under dataset/ you will find the following:

• fda/ - FDA-approved small molecules with ~1.5K molecules
• riniker/ - Riniker charge sets with ChEMBL (40K) and ZINC (90K) subsets
• pdb/ - the PDB Ligand Expo containing ~23K molecules from the PDB
• nci250k/ - The NCI250K set of small organic molecules of interest to drug discovery (~120K)

charge-datasets's Issues

Duplicate molecules in the SPICE dataset

There appear to be some duplicate molecules in the SPICE dataset.

Some molecules are uncharged

There appears to be an issue with the SPICE dataset where some ligands have all zero charges, suggesting that charging failed but the molecule was still included in the dataset. We should fix this.