2022-08-31 Wagner/Mitchell Check-in meeting notes

Participants

  • @Jeffrey Wagner

  • @Josh Mitchell

Discussion topics

Item

Notes

Item

Notes

General updates

  • JW – Toolkit 0.11.0 release cut, conda package made. About to make release announcement

    • JW – Could you update bespokefit for Toolkit 0.11.0?

    • JM – Should I also make a new build of the last version with a downpin?

    • JW – It looks like it’s already downpinned

    • JM will update bespokefit to pass with latest OpenFF toolkit and cut a new release

  • JW – Toolkit showcase PR review

    • JM – I like all the changes except the box size change.

    • JW – I think we need to keep the box size small, until we get the total runtime under 5/10ish minutes.

    • JM – Ok, I’ll review this and push any changes I want directly to the PR branch.

    • JM – I may take a smaller protein-ligand system too

      • JW – Consider grabbing a complex form the protein-ligand benchmark repo (or PR#52 which is re-preparing a lot of targets and adding a few more).

      • JM – Great, I’ll check that out, and will look for interesting chemistry in the ligand.

  • JW – Modifed AA notebook --> Loading unusual AAs? (How could we do this?)

    • JM – Maybe expose private API for people who want to do this, but refer to it as “experimental” so folks have the proper expectations.

      • Would need to modify either the substructure yaml, or access it during runtime.

      • JM – Couldn’t we just use SMARTS?

      • JW – I think it’s more complicated. We discussed the needs for our substructure format in an earlier meeting with Connor Davel. We could use something liek SMARTS but would need to have some restrictions on contents and some special meaning for atom indices. Alternatively we could just let people feed in new info as cif files.

      • JM – Could be good to let people feed in additional substructures as cif.

  • Bespokefit PR feedback

    • (Reviewed for a while)

    • JM – Ok, I think it should be straightforward to dump the structures when there’s a failure, and I can just use the expected connectivity from the qcschema mol. I’m somewhat ambivalent about putting the check methods on the class. I’ve responded on the PR and will make those changes.

    • (JM posted a response, will follow up)

Todos

  1. (high) Resolve feedback on BespokeFit PR and make discussed change

  2. (medium) Prepare material for Bespokefit workshop (materials from last time)

  3. (medium) Update bespokefit to work with Toolkit 0.11.0

  4. (medium) Centralized examples (from)

    1. (high) bespokefit protein-ligand or ligand-in-water example

  5. (medium) Centralized OpenFF “book” docs

  6. (medium) Toolkit docs cleanup

  7. (medium) Toolkit revised user guide (+-unifying/centralizing package user guides)

  8. (medium) Come up with “milestones” for making first of three videos this year (like, “first video on covid spike protein and small molecule, filmed this day, edited in that range, etc…”)

  9. (medium) bibtex blocks for website “how to cite” page on

  10. (low) Propose policy for using GH citation machinery

  11. (low) Check main website for broken links

  12. (low) Conda env yamls for each release (automated inside of Toolkit’s single-file-installer action)



Action items

Decisions