2021-03-01 Developers coffee Meeting notes

Updates

• SB –

  • Thinking about WBO stuff – Parity between OE and AmberTools

  • Also working on graph convolutional net implementations for partial charges and WBOs

• MT –

  • Org work/interoperability

  • Opened + merged a bunch of toolkit PRs

  • Some discussion on interfacing with legacy force fields

  • Started looking at strawman repo

  • Fixed constraints in openmm export

  • Met with biosimspace

  • Continuing with conversion/energy matching tests

• DD –

  • Improved unit tests on openff-benchmark optimize

  • Focusing this week on getting new release out

    • Need to decide what to do with Swope’s iodine message

  • Going to see many partners finish calculations this week

  • QCA – Added use of conda caching to error jobs. Looking at other areas where we can add new features.

  • PEPCONF – PB and TG have discovered many useful things that will improve our success rates on hard molecules. Need to figure out how to implement them.

  • Will work with Xavier Lucas on ANI torsiondrives – Will plan on adding this to openff-benchmark

• PB –

  • WBO fitting experiments and benchmarking

  • Tried to submit another set of genentech molecules, hit some validation errors. Working with JH to resolve.

  • Merged QCSchema roundtrip – Had to fix some test decorators.

  • Will there be a new forcebalance release to prepare for the openff.toolkit move?

    • JW – Yes, I’ll help bug Lee Ping about this. Also the 0.9.1 release will have an equivalent, one-time-only 0.8.4 that has all the same features. 

• JW –

  • Made a reproducing test+bugfix PR for WBO conformer dependence thing

  • Thanks SB for work on biopolymer SMARTS matching. I'm working with Connor Davel of Boulder to do more performance profiling

  • Benchmarking – Prepared mirror image deduplication PR (will defer to season 2). Also opened PR to make stereochemistry errors on export treat molecules as error_mols.

  • Spent a lot of time/energy on communicating with Andrew Dalke. Could help us with big molecule representation and toolkit interoperability issues.

    • MT – Our development practices are particular. It would be good to include our PR flow in the contract to ensure that work is done appropriately in our ecosystem.

•  

0.9.1 release

• We will cut 0.9.1 this week

• JW will back-port PRs to 0.8.4rc branch

  • Plan of action:

    • JW will TRY to back-port all PRs, but if there’s any smoke, he’ll just back-port PRs that are necessary for benchmarking (check with SB on whether any current fitting experiments are using 0.8.X?)

    • JW will ping MT to do a “smoke check” before merging 0.8.4 cherry picked commits.

  • Release process will have some additional steps/constraints:

    • BEFORE cutting 0.9.1

      • 0.8.4 should be cut

    • AFTER cutting 0.9.1

      • Benchmarking will need to update production environments to get off omnia/label/rc and pull stable

      • In release announcement, say that this is the last release to support python 3.6