MT – I’ll be offline most of this week, meetings cancelled.
JM is offline this week
Individual updates
DN
Started working on deadlines for different projects. Using a tool/technqiue called “critical path”, which will indicate our biggest project liabilities. One thing I’ll need to know is how much time each item will take to be developed. This will be tricky since there’s a lot of unknowns in our processes. In my previous roles this had been a bit easier since I know how long building woudl take to build, how long to hire, etc. But this will be different since it’s a science/technology project. So one thing I want to ask is whether we have deadlines for items and which calendar we follow.
JW – Pharma renewals are in october, that could be interpreted as our “year end”. But mostly we’ve worked without deadlines, and just used target dates for FFs to justify priorities.
MT:
Bespokefit single installers done (RDKit+AT+xTB)
Some other improvements to BespokeFit - eg tests had a hard dep on OE, now it’s soft.
Interchange now has feature parity with the toolkit’s OpenMM export except for GBSA and custom plugins (both pushed to releases after 0.11.0/0.2.0). Need to re-run regression tests but will likely call a meeting soon to give an update to stakeholders.
Re-ran value propagation, multi-molecule, and WBO tests, all passing.
Single-molecule tests running now, will update on results in a couple of days.
We should drop openmmforcefields from toolkit examples, common cause of breakages/incompatibilities and not a project we have control over.
JW – I don’t think we’ll use espaloma in production code. Will instead use nagl+gnn-charge-models. I made a parameterhandler for expaloma in smirnoff_plugins but that’s more of a proof of concept. For openmmforcefields I think we should remove it from our “first class” examples, and maybe keep it just a step above”deprecated”. SB is migrating the repos he’s comfortable with migrating to the openforcefield org, and encouraging us to fork the others when we need to further develop them.
DN – Thanks for looking ahead to potential issues and being proactive in problems that affect the whole ecosystem. I’m really happy that people on this team are motivated to solve our big problems.
DD
will be offline Wednesday - Monday, assuming can attend conference
fah-alchemy
finished gufe#13 with Richard Gowers on Friday; this included a rework on how DAGs are represented, and the PR as a whole defines the data model for Transformations, Protocols, and AlchemicalNetworks
meeting with OpenFE folks on next steps from their perspective after this call; will coordinate where possible on ResultStore
will need to move on to building initial set of Protocols, Executor, to target user stories
worked on components of protein-ligand-benchmark; finishing out my assigned PRs this week.
QCArchive
worked with Ben last week on openff-qcsubmit; completed work on QCSubmit Datasets, now working on result collections
Ben has agreed to take this on directly, since this helps him to refine the API points on QCPortal - Will be mutually advantageous since BP gets to see use cases and is an expert in the API changes + performance
worked on MPI cluster deployment; major limitation is 10GiB scratch space, which precludes our use for e.g. SPICE, but not our usual method/basis choice
PB: Last two weeks
Worked on Sage manuscript, getting close to first draft but still seems far away.
Wrote blogpost on qca-pipeline, sent to David for review.
JW – It’s fine for blog posts to have mistakes and typos, so don’t stress too much about it
PB – No rush on reviewing, DD.
Getting back to fitting experiments and pushing towards a point release, I have most of the iterations before, documenting them properly and decoupling a mix of improvements:
Made a list of missing chemistries (6 parameters), which may not be super relevant, so current solution is to include them in some of the general parameters by making changes such as making a single bond to be flexible or adding an extra valence, we already have and not bother about them much. There are molecules with those chemistries on pubchem though.
Looking at problem chemistries now.
General qca monitoring and setting up qm-managers.
JW
Bespokefit cleanup+release (Thanks MT for hardening and SFIs!)+support
DD – Is xTB as good/equivalent to psi4?
JW – We’re not sure, JH suspects that it is but we haven’t tested yet. The quickstart guide now uses xTB for technical reasons, but I added a disclaimer that it’s not the same was what we’ve shown in earlier results.
Prepared to put cross-repo examples into openff-docs (JM will lead this charge, will ultimately include examples like “make parameters with bespokefit, prepare protein using PDBFixer, combine systems in interchange, run calcs using OpenFE software”)
User support, specifically pfizer. Tracked it down to a weird OE license issue, redirected to support@OE
Mostly clearing backlog of comms, org stuff, PRs
This week I'll be charting the path forward for the 0.11.0 release