2021-05-26 Dalke/Wagner Task Speccing

First task/practice with OpenFF development

• New molecule test set?

  • MiniDrugBank has untraceable provenance, suffers from openbabel conversion, many other problems

  • Would like a new test set of ~100 to ~300 organic (C, H, O , N, P, S) molecules to ensure equivalence between our use of OpenEye and RDKit.

    • Provenance of this set must be recorded (so, either keep names from original sources and record version of original datasets, or keep script for filtering them)

    • In the short run, we should mark+skip “bad” entries in this dataset, and use them as guidance for future changes

    • This file should be SDF, and should represent valid SDFs grom various sources.

• Alternate: Fix R<n> interpretations between OE and RDKit

  • Would start by opening a PR or issue with a test case showing nonequivalent behavior between the toolkits, and then proposing fixes

  • AD – Could be a good fix to offer a warning when users might be burned by this behavior

• For next meeting:

  • AD will rebuild dev environment: https://open-forcefield-toolkit.readthedocs.io/en/latest/developing.html#setting-up-a-development-environment

  • JW will review AD’s PR

  •