2022-03-09 All-hands meeting notes

OpenFF toolkit 0.10.3

• Release made, fixes issue where TIP3P parameters weren’t being assigned correctly. An incidental change broke forcebalance. So we made a new FB release (v1.9.3) yesterday that should fix it.

QCArchive storage resolved

• QCArchive was running out of storage space, largely due to wavefunctions saved with initial SPICE set submission, and also optimizations and torsiondrives storing Wiberg and Meyer info for every optimization trajectory step. BP asked us what could be deleted, and did a risky deletion operation to reduce storage utilization from >90% of total QCA capacity down to ~50%. We’ll be implementing new defaults in the future to have datasets require less space and only store what they need.

• Separately, the qca-dataset-submission repo was getting too large and cumbersome, and David Dotson switched it over to use git lfs. Folks with existing cloned versions of qca-dataset-submission will need to re-clone using the new instructions in the repo README

Bespokefit release and status

• Release coming soon

  • ForceBalance and Toolkit incompatibility, we may need you to install with version pins.

  • Critical bugfix needed, will try to make a release tomorrow - Psi4 ncores propagation and simpler QM spec option.

Charge fitting datasets

• Generating some ESP datasets on QCArchive. Training set made, using industry benchmark set as a test set. So these should be ready for charge fitting work.

  • Training:

  • Test: ~10k molecules, ELF10 conformers for each

OpenFE update

• Notebook OpenFE has been working on which demonstrates the current state of the toolkit and how it fits within an RBFE calculation using OpenMM:

  https://github.com/OpenFreeEnergy/ExampleNotebooks/blob/master/openmm-rbfe/OpenFE_showcase_1_RBFE_of_T4lysozyme.ipynb

• An environment.yaml file to build the required conda environment can be found here: https://github.com/OpenFreeEnergy/ExampleNotebooks/blob/master/.binder/environment.yml

• A binder instance can be obtained here (please don't run the MD steps on binder, their free architecture cannot support CPU intensive simulations): https://mybinder.org/v2/gh/OpenFreeEnergy/ExampleNotebooks/HEAD

• You can install everything you need with conda env create openfe/openfe-notebooks. That will give you the correct environment to play with the notebooks.

• Any feedback would be immensely valuable.

F@H project update

• Note! Per the project plan, this is not a formal place to discuss this project or make decisions! Please attend the Tuesday meetings if possible (or contact JW, the OpenFF Approver for the project) if you suggest modifications.

• User stories were being collected on the issue tracker on fah-alchemy

• OpenFE has accepted approver status