2022-10-04 Meeting notes

MT – I think we should use QCF. Wasn’t sure before, but I tried out the QCF/Submit stack and it seems solid.

MT – Season 1 benchmarking used a lot of rdmols. Do we want to continue that pattern? How much openeye should be in this? Or should we use OFFMols?

• JW – I think we picked RDKit because

  • it let us handle molecules even if they weren’t valid OFFMols

  • it exposes lots of knobs for conf gen/rmsd calculation/alignment

  • everyone has access to it and it’s open source

• LW – RDKit offers a lot more knobs than OpenFF (eg. max attempts)

• MT – Do we have a preference for the backend in this case? OE is easier to work with and more performant.

• LW – From a sustainability perspective, RDKit is more promising. I see the point about the RDKit API being somewhat “sticky”, so wherever possible it’d be good to route calls through OpenFF toolkit.

• MT – I didn’t expect that. I looked at the code a bit and saw a lot of direct calls to RDKit.

• JW – Routing through toolkit API may make initial implementations traightforward and more reproducible, but would make changes more expensive. This would also make it possible for OFFTK releases to change benchmarking results.

• MT – An RDKit release will always make it possible for the conformers to change. Maybe we’re talking about two things -

  • Should core representation be OFFMols or RDMols?

  • Should we “talk” to cheminf packages directly or through toolkitwrappers.

• MT – Do we agree with “don’t directly call openeye”?

  • LW + JW – Yes

• MT – Should we ever call Openeye?

  • (undecided)

LW – A while ago, JW asked if I need Recharge to be updated to the new toolkit. I siad “no” at the time, but I was wondering if there’s an ETA.

• JW – MT and I went over this in prioritizing earlier today… Unlikely before Dec.

• MT – There are a few things that need updating - BCCs, Vsites …

• LW – I need to use the classes from Recharge. I’m able to copy and paste stuff to get things running. But it’s ugly and I’d prefer to use a stable release.

•  

• MT – I can do this in the next few weeks if we don’t handle vsites

• LW – That works for me

• JW – That would be great - Please go ahead with these fixes and a new release

• (We changed recharge ownership on openff/status to matt)

(General) We should have the organization be more deliberate about data/software “warranties” or “lifetimes” - How long do we support running old versions of datasets? Is there a difference for academic vs. industry facing? etc.