2022-08-29 Core Developers meeting notes

General updates

• JW – LW is offline T+W (maybe M+T actually, given date line). Shouldn’t affect any calendar meetings.

• JW – 0.11.0 release very likely today

• PB – Offline next week

• CC – Offline Thurs+Friday this week.

Individual updates

• DN

  • Email from Katharina, feedback on annual report. She’s happy with the annual report, but had some feedback on the claim that governing board asked for biopolymer FF, so she may want a revised version put out. Will discuss with JW later today.

  • MT and CC were on the FF release call when we were kind of invited to participate on the science roadmap building with PIs. Discussion about whether it would be best to have two different roadmaps that would eb aware of each other. JChodera suggested running micro-workshops where ad board members give short updates on their biggest failures and biggest wants. So I invited the ad board to participate and give feedback - KMeier is positive on this. We’ll invite the entire team to these meetings (will be in the ad board time slot).

    • JW – I really like this idea.

    • DN – Yeah, this is a logical followup from the ad board not insisting on a Sage 2.1.0 release.

  • MT – Re: roadmaps - There seems to be two axes - (academic <--> industry) and (science <--> infrastructure) - I’m not sure which roadmaps are being proposed, but I’d be cautious about assuming significant overlap. So this seems like a tricky area to plan around.

    • DN – A lot of what was discussed was “long term science”, which seems pretty academic. But the short term science seems like it’s of great interest to the industry-supported/consortium roadmap.

    • JW – I’d like the idea of consortium and initiative (academic and industry) roadmaps. But I’m happy to merge infrastructure and science on both sides

    • DD – It seems like it may be ideal to have a single roadmap, with things labeled as “consortium” vs. “initiative”.

    • JW – I’d like us to at least start with separate roadmaps, but we can regularly meet to discuss helping on specific items and making hard commitments to complete tasks/deliverables.

    • DN – This kinda arises from a need to show that NIH money isn’t going to non-NIH goals, and likewise for industry support.

    • DD – I think this could be accomplished by having a column indicating which items are in consortium vs. initiative.

    • DN – I don’t think the initiative roadmap would be too complicated - I think it will be a few items that are longer-term. But the PIs will need to handle the initial prioritization of their tasks before we discuss shared goals.

• MT

  • Flurry of smaller patches into Interchange Iterating with JMitchell and JWagner.

  • Working on toolkit around virtual sites. OFF toolkit TIP5P is slightly different from OpenMM TIP5P, we did the math by hand and OFF toolkit positions are correct, OpenMM positions are different. Took a long time so we didn’t look further.

    • JW – We’re getting different positions from OpenMM’s TIP5P. Not sure why, doing the math by hand it looks like we’re getting the right geometry. We’re going to stop digging.

  • Documenting release.

• DD

  • Protein-ligand benchmarks

    • finished rework of gufe DAG system to make consistent with the rest of the library; waiting for review from OpenFE

    • stood up development infrastructure on Chodera Lab AWS: small host with folding@home work server, neo4j

      • JW – neo4j license - Compatible with OFF licenses?

      • DD – I think so - I checked this weekend whether we need anything from enterprise edition and I don’t think we do. It’s installed at the apt-get level so we should be able to deploy it on lots of hosts. So it’s not conda installable but it should be quite straightforward in our deployments. This also isn’t specific to F@H, so OpenFE should be able to reuse it with no trouble.

    • started up work with Ivan, Mike on nonequilibrium cycling protocol in perses using new gufe DAG system

    • now performing development cycles on roundtripping AlchemicalNetwork structures into neo4j's data model in fah-alchemy. Trying to find roadblocks early in case we need to find another solution. But it's looking good so far.

    • will be working on service API and compute (via dask) later this week in parallel

  • QCArchive

    • submitted SPICE openff-default sets prepared by Pavan; many appear to have hit systematic errors and are mostly complete

• PB

  • Spent sometime looking at the phase angles of torsion parameters that have 90/270, which are odd since phase angles are usually 0/180, and there seems to be no need for them. Chris Bayly didn't put them in but Daniel McKay might have some insights about their origin.

  • Was looking at the torsion multiplicity fitting, didn't succeed in the first attempt, some angles (a11, a14) optimized to weird values.

  • Wrapping up incomplete sections in Sage manuscript.

  • Global benchmarking metrics might be deceiving, especially RMSD metric, and there is a need to look at bad chemistries in a tiered approach, with the angles and bonds coming first and then torsions. Found instances of bad phosphorous angles in some of my experimental fits, and the sulfonamide issue with Sage.

• CC

  • Still ping-ponging revision of LiveCoMS review with MGilson. Focusing on NMR section.

  • Progress on preparing protein parameter fits based on template from PB

    • qcsubmit fails for TorsionDrives with multi-dimensional grid indices. Opened an issue on qcsubmit here and made a patch in a local branch so this wasn’t a blocker.

      • JW – Let’s leave this in the branch for now, could you push it to GH and name it in the issue so we can dig it up later?

      • CC – Can do

    • Baker-Hubbard hydrogen bond criteria filters 72 protein TDs down to 17 TDs. Skipping this filter for protein TD datasets so that we don’t throw away 75 % of training data.

    • This week setting up ForceBalance runs on UCSD cluster

• JW –

  • Getting examples tests running

    • MT – The issue is that the interpreter won’t import PDBFixer. Has anyone experience this before?

    • DD – OpenMM’s PDBFixer?

      • MT – Yes.

      • DD - There’s a tool for SSHing into a live GHA runner, useful for debugging. It’s called tmate:

  • Finished icode handling

  • Worked with MThompson on vsite verification

  • Worked with JMitchell on toolkit showcase refactor, migration guide/changelog finalization, and performance improvements

  • Doing final checks on examples.