2023-11-28 alchemiscale : dev group meeting notes

• JC : OpenMM 8.1 is now released

  • so if you’re creating new envs, likely to pull this; probably no need to pin, but FYI if you do run into issues

  • MH : performed tests with it, didn’t observe any issues

  • IP : ran perses benchmarks with it, no observed regressions

  •  

2024 development strategy

• JC : being able to run useful and short tasks for adaptivity is key for us

  • being able to rapidly iterate on network generation, including experimental structural data

  • (near) real-time free energy calculations exported to external systems (e.g. medchemists views)

• JC : next step in tech tree could be

  • adapting effort where needed could be glued into surrogates

    • JC : a model to estimate difficulty of FEC

      • could then be used

      • could also have models that give out affinity estimates

      • could train models on large networks as well

    • could be bolted on to alchemiscale as a separate service

DD development update

IP : protein-ligand benchmarks

• IA : needed a stable point for OpenFF to build on

  • hitting issue with partial charge generation; fixed them with conformer generation, but now hitting new problems with antechamber

    • sqm is returning irreproducable results in specific ligands, most ligands in p38

    • answer here is to drop antechamber

    • looking to change protocols for charge models

    • JC : looking to retrain espaloma-charge by end of December

      • IA : been testing out nagl charges; keen to try espaloma charges too

      • JC : both espaloma and nagl use same model; looking to use a larger dataset

• DD : JC presented a vision for how to build automated process for protein-ligand benchmarks

  • socialized further?

  • JC : no initial driver for this at the moment, but some activities to prove the concept for a few systems of interest

    • may be able to show that this can be done sourcing from BindingDB, spruce them, etc.

• DD : is there a regular meeting for protein-ligand benchmarks working group?

  • IP : not yet, should probably start one up in January

  • JC : will contact current folks involved, see if we can get participation in v3 (automated pipeline)

• IP : perses-openfe interface

  • ready for review: openfe 0.14 support in feflow:

  • want to change namespace for feflow to be in openfe Github org

    • DD : can you add my Github user (dotsdl) and Ian (ianmkenney)

    • IP : currently can’t add non-admin

      • DD : can we set choderalab org base role to Write

  • IP : meeting next Monday, making forward progress on this board: Perses-OpenFE roadmap • choderalab

  • DD : need any additional resources for executing?

    • IP : IA, need help with charge issues?

    • IA : should be good to go