2021-05-14 QCA User Group meeting notes

MolSSI - Public QCArchive

Ben

• Development on next branch will slow over next 6 months

• Portal release probably needed soon

  • DD: have one issue encountered Wednesday with Lorenzo in the client; will investigate and try the band-aid of making the client code try both, or change the REST model to be stricter

    • find ID, share with Ben

• Working on the renewal; thank you for everyone who added a blurb for how QCArchive enabled new science

  • https://docs.google.com/document/d/1WMc2Lw65cLbxbGVsFredWmuYe_Aq9XrbbNszxflhfD8/edit#

• JW: MolSSI educational resources in particular are another area of high value for this community, and OpenFF in particular

User questions / issues

• DD: Simon, anything you need for industry benchmark dataset?

• SB: no, I think these are just bugs on the toolkit side; nothing needed at this time on the QC* side

• JW: toolkit behaves differently depending on when it sees mapped smiles vs. unmapped smiles

• TG: noticed that the server availability appears to have decreased over the last few months; appears down for 10 - 20 minutes

  • BP: the public Fractal Server?

  • TG: yes, next time I get a ‘connection refused’ will reach out to see if we have corresponding activity in the public server logs

• DD: Gen3 torsion set, technically ready to go

  • SB: haven’t had a chance to give it a look; let’s wait to submit this so we can give the industry benchmark set more time to complete

• DD: Gen3 Opt set in preparation by Hyesu; where does this fit in priority?

  • SB: keep industry benchmark on top for now

• JW: connectivity filter discussion: https://github.com/openforcefield/openff-toolkit/issues/936

  • SB: the filter I shared in the DM could be used to identify the cases in the dataset impacted by this issue

  • JH: course of action:

    • keep current dataset going

    • will have a QCSubmit download filter that handles cases like this

    • should also introduce a fix in the toolkit that throws errors when we encounter cases like this (basically, a validation error)

  • JW: believe we need to:

    • ensure that validation in openff-benchmark captures these cases, filters them out

      • probably requires toolkit fix

    • JH: for the QCSubmit filter (for export), can just look at the CMILES for any implicit hydrogens when hydrogens on the molecule are otherwise explicit

    • JW: action items

      • resolve toolkit issue

      • raise issue on openff-benchmark that CMILES such as this should be filtered out by validation component

    • JH: will go through dataset and try to get all impacted IDs

      • will inform the approach for the filter