2023-12-21 Mitchell/Wagner Check-in meeting notes

General updates

• JW –

  • This will be the last meeting of the year

  • Thanks for ping-ponging PRs with Matt and helping with user support.

  • Lead team says don’t do Sage lipids, so either:

    • plaster it with warnings about how this isn’t a good idea

    • Use Sage for some general other mol (cholesterol or something?)

    • Somehow include Sage in an RNA+lipid sim

      • (should loop in MShirts if you go down this road since his lab is doing some exploratory work in this direction, but hold tight to scope)

• JM – Tried protein prep stuff. Seems to have worked, but OFFTK doesn’t like loading the PDB, and that’s as far as I got.

  • JM – MHenry said that his funders didn’t want protein prep, so he doesn’t have material on that.

  • JM – And I’m skeptical that OpenFold can do loops, haven’t reached out to JWei

    • JW – it’d be really impressive if we could even kinda land this

    • JM – I doubt it would do a good job on loops.

    • JW – Maybe we could find a case where there’s something structured but unresolved? But OpenFold deployment is still a bit rough so maybe this would be best as an appendix.

•  

•  

•  

•  

Trello

• JM – OpenFF-docs serves basically as deployment tests for the ecosystem packages (though it allows pinning packages, and currently things ARE pinned). It’s been broken for a few weeks and I need to delve into it, and it’s complicated figuring out whose jurisdiction this is.

  • JW – These could be really useful. I’d like to see if there’s a way to spread out the work and responsibility for keeping these tests green over Matt and I as well.

  • JM – I don’t want to be responsible for everyone’s examples - It’s not always dependency conflicts or pinning, sometimes it’s just squarely within the repo.

  • JW – Oh, yeah, I agree.

  • JM – Could endorse a single “holy environment” in install docs

  • JW – Skeptical for 3 reasons:

    • I want to be sure it works+is maintainable before we change messaging to users

    • There will always be capability compromises/dependency stack forks (eg currently with qcsubmit and new QCF)

    • (something else)

    • JW – So, not against it, but would want to try sharing responsibility for this for a few months to see how it works out.

  • JM – Also, the docs ecosystem doesn’t currently have qcsubmit+psi4+bespokefit. So that would increase complexity (I tried this recently and hit some sort of issue with psi4)

  • …

  • JW – So, maybe we should have a rotating “openff-docs hat”, where if openff-docs tests turn red while you’re wearing the hat, you have to figure out whether it’s a packaging issue and present the options to resolve, or contact the maintainer whose example broke and tell them to fix it.

  •  

Workshop planning

https://openforcefield.atlassian.net/wiki/spaces/COMMS/pages/2719055920

• JW – DM mentioned that the workshop should teach people when something is unsimulatable (like, tell them not to do heme)

1. (high) Look into doing SMIRNOFF workshop with RNA (openmmforcefields) + lipid (for example, mRNA vaccine formulation compound DSPC) some sort of modified cholesterol (Sage+bespokefit)(details here)

2. (high) Review protonation and tautomer state PRs

3. (medium) Prototype “prepare any PDB” workflow”.

   1. JM will contact MHenry to coordinate with his Bristol Meyers Squibb simulation setup.

   2. Find+categorize cases where structures aren’t suitable (eg heme groups, metals, other weird cofactors)

   3. (optionally) also reach out to JWei to incorporate OpenFold for loop/missing residue modeling.

4. (medium) Look into clean options for using NAGL to charge NCAAs (can it handle full proteins in a reasonable amount of time?)

5. (medium) Fix openff-docs CI/figure out what’s up with deployments

6. (low) Reach out to science team for additional vignette ideas

7. (low) Propagate PDB-bond loading options into nglview