2020-05-06 BespokeFit meeting notes

JH – QCSubmit is nearly ready to go

• DM – It would be good to submit the Bayer set with some other settings

• JW – @David Cerutti (Deactivated) will be submitting a lot of amino acids soon, I’ll instruct him to contact @Joshua Horton

JC – HX, you can check out the status on SiliconTXtorsionbehcmarkset1 on QCPortal

JH – Currently I assume that molecule is correctly formed/specified

• DM – This overlaps with another need for an OFFTK which would be a molecule checker that highlights problems and/or tries common sense things to fix it.

• JH – I’ve opened an issue

JH – Worked together with Matt Thompson on quality checks in QCSubmit

• JC – should checks live in QCSubmit or OFFTK?

• JW – While it needs additional dependencies and iterates quickly, we decided to keep it separate

JH – Running OpenMM calculations on QCA will require releases of OpenMM, OpenFF Toolkit, QCEngine

• JC – I’ll push on OpenMM

• JW – I’ll schedule to talk about this in QCF meeting

We could make a command line interface to bespoke workflow

• We should standardize on CLI practices within OpenFF

• https://openforcefieldgroup.slack.com/archives/C8NE3J96U/p1588782195081800
  @Josh Horton @Jeffrey Wagner: Here is docopt for constructing CLIs: http://docopt.org/
  Here’s an example of how we use it in YANK:

  • dispatch: yank/Yank/cli.py at master · choderalab/yank

  • This allows any module in this directory to be automatically dispatched: yank/Yank/commands at master · choderalab/yank

JW – We’ll want to be very deliberate with how we build out infrastructure. If we’re not careful, we’ll wind up with Josh doing infrastructure support.

• @David Dotson can help with setup and documentation for local QCA instances

• Galileo/hypernetlabs people could offer commercial hosting and provide support for larger-scale deployments

DM – It’d be interesting to see what fragmentation behavior is – Like, some preliminary data on how many fragments we get from fragmenting common kinase inhibitors may be instructive

JH – Conformer generation seems to take a long time

• JC – There should be ways to reduce the computational cost of OE conformer generation

•