2020-10-09 QCA Submission Meeting notes

Hanging INCOMPLETEs

Trevor

• Is it necessary to take any action on a TorsionDrive that had a hanging INCOMPLETE that was regenerated?

  • BP: not certain, but should be fine

    • Ben will check the code path to make certain

Protein Fragments Optimization

Josh + Trevor

• TG: what’s the best way to review some of these?

  • DD: I tend to start with intent from the README/Jupyter Notebook, then make sure intent is encoded; not sure if that’s helpful

  • TG: Mean more what QCSubmit doesn’t do that we should be doing?

• TG: Might be good to specify if we want a valence constraint or a non-valence constraint in the QCSubmit dataset object so validation could be done downstream.

• TG will review and note anything DD missed

MM/ML dataset

David

• TG: can we create a coverage report from QCSubmit’s coverage analysis to show which parameters are covered by a TorsionDrive submission?

  • TG: I’ll give this a play; could be an additional dropdown or separate report, or also an artifact that we commit (like the torsion PDF)

STANDARDSv3

Trevor

• DD: would like to establish a target date for adoption

• TG: Mid-November would be good to get it in before the onslought of fitting datasets

  • will ping for review once draft is up

QCSubmit advancements

Josh + Trevor

• TG: will be able to submit much, much faster with a multithreaded approach

• TG: every step in the workflow has parallelization over molecules with multiple processes

  • speedups make QCS usable beyond 500 molecules

INCHI keys

Trevor

• Under what conditions do INCHI keys collide?

  • Josh: doesn’t encode tautomers

  • QCS now deduplicates in a modified INCHI key, which handles tautomers

TorsionDrive selection by SMARTS

David

• Preparing this for users; is there already some machinery in QCSubmit?

  • TG: coverage already there as a filter

  • JH: SMARTs filter already there

  • DD: will start with those!

PCM-based implicit solvent

Trevor

• BP: Lori’s working on if you ask for PCM, you get C1

• JH: Going forward we can force C1, once the above we could also get the same for older submissions if desired

• Conclusion:

  • JH: check that C1, PCM in the molecule; if not, inject it

  • TG: will include in STANDARDSv3

ESPs and wavefunction storage

Josh

• We can store more wavefunction properties now; DGS fixed a bug in psi4 blocking this

• TG: is there a release including these?

• JH: will try to find which one includes this

  • will find the issue blocking wavefunctions->ESPs calculation

• Would do wavefunctions as separate BasicDataset submissions

  • would need to get attributes on them

  • DD: would need to be a change in qcfractal; will take on

• TG: do contributed values factor in nicely here?

  • BP: there is some interest in modifying datasets to have extra metadata

  • TG+JH: would be nice if datasets have a consistent interface with rest of datasets with entry.attributes.cmiles