2024-07-23 alchemiscale : dev group meeting notes

JC – OpenADMET will be funded

• DD – Congrats, how will the tooling we’ve build here contribute?

• JC – They’ll be using OpenFE, leveraging OpenFF and alchemiscale. But most of the efforts will be around structure-based machine learning, rather than explicit FE simulation.

DD : alchemiscale roadmap

• Q1 : complete “living networks” performance improvements

• Q1 : Folding@Home compute services deployed in production

  • finish MVP, with integration test suite by 2024.03 2024.06

  • DD – alchemiscale-fah #7 (massive PR) merged with help from Hugo and Ian.

  • perform FAH tests with volunteers during 2024.04 2024.06 2024.07

    • public work server up by 2024.03.15 2024.06.11 2024.07.19 2024.07.31

      • DD – Will begin doing test runs at this time.

      • JW – Excited to see how this goes.

      • DD – I’m hoping that this will uncover performance bottlenecks, and volunteers will be unhappy about points awarded. Points need to be assigned on a per-project basis, which may be tricky. Talked to Joe Cochland about how points could be assigned - we thought about assigning them per run instead of per project.

    • confidential work server up by 2024.04.01 2024.07.01 2024.07.31 2024.08.16

• Q2 Q3 : develop Strategy structure, initial implementations

  • aiming to begin design 6/26 7/10 7/24 sprint, followed by MVP development during July August

  • DD – Alchemiscale #184 - Assigned IK to lead this effort.

• Q3 : enable automated Strategy execution by end of Q3, 2024 (2024.10.01) mid Q4, 2024 (2024.11.15)

IP: feflow needs

• IP – Hit a blocker in PR #49 about how to execute units. Code error is that it expects two positional arguments. SetupUnit is a PRotocolUnit, I want to execute just that without needing to run the whole protocol.

• IA – We do protocol.create, then get a DAG out, then get the unit we want.

• IP – I want to avoid that, since then we need to make and run a whole protocol to get the result out.

• IA – You can run individual steps of a protocol.

• JC – So, there’s a simpler approach where you replace the units that would run locally with something that runs of F@H. Like,, the current paradigm is that … instead we could change it so that things that COULD run on F@H are substituted out…

• DD – That’s how the noneqcyc protocol works. It has F@H specific compoents that are substituted into the normal protocol.

• IA (chat) – this is our usual approach: https://github.com/OpenFreeEnergy/openfe/blob/cc5ed8c985fd386c9eedc545a24fd192db405e63/openfe/tests/protocols/test_openmm_equil_rfe_protocols.py#L196-L205

• DD – We probably have tests of protocolunits in GUFE (I use their test suite as a reference a lot)

• IP – Wanted to avoid creating the whole DAG, it makes it less of a unit test (and this is already not really a unit test)

• DD – We’ll return to this at the end of the meeting.

•  

alchemiscale development : new sprint spanning 7/24 - 8/5

• cutting alchemiscale release 0.5.0, deploy to alchemiscale.org, including openfe + gufe 1.0:

• DD – Wanted to run some F@H testing on this before it comes out in case major changes for compatibility are found.

  • (General) – 3.12 may be tricky - people vaguely recall upstreams being unavailable but can’t name a specific one

  • DD – I’ll try adding 3.12 to the matrix and see if it fails, can drop to 3.11 if needed.

• IA – Are you saying there will be an alchemiscale release this week-ish? I made plans to do an OpenFE release in the next few weeks.

  • DD – If it’s an incremental release, we can easily make a new build.

  • IA – Yes, shouldn’t be API breaking, will just let us do better solvent boxes.

  •  

• architecture overview : https://drive.google.com/file/d/1ZA-zuqrhKSlYBEiAIqxwNaHXvgJdlOkT/view?usp=share_link

• coordination board : alchemiscale : Phase 3 - Folding@Home, new features, optimizations, targeted refactors

  • (see recording ~20 mins)

  •