2023-04-21 Thompson/Wagner Check-in meeting notes

Participants

  • @Matt Thompson

  • @Jeffrey Wagner

Discussion topics

Item

Notes

Item

Notes

General updates

  • JW – We’re looking at annual meeting talk - It’ll be kludgy if we’re switching speakers every few minutes - Are there specific things on Interchange or Benchmarking you’d like to talk about, or would you be happy helping prepare slides but not presenting on the day? We’d love to show heaps of videos of cool simulation “vignettes” to get people excited, and if you wanted to grind through those that could be very impressive to point to a GIST and be like “yeah, all of this is real, go wild”

    • MT – I might not agree that “reading docs/releasenotes” is a bad use of time. My experience actually being in meetings with potential users has effectively been “reading docs”. So from my persepctive I don’t think I can fill 20 minutes talking about interchange, since the most advanced thing you can do it solvate a ligand and export to OMM/AMB/GRO. This is basically the extent of capabilities until Rosemary (unless we fall back to ff14sb). Most of the other flashy stuff is not in a state where I’d like to present it. Instead the big stuff I’ve tested a lot is like alterntive functional forms and stuff Meghan is testing over the weekends. Really rock-solid stuff is serialization… kinda vectorized representations.

      • Direct parameter fit

      • Multi-format export

      • Serialization…

    • JW – Would you be up to not speak, but help with slide prep and vignette creation?

    • MT – I’m happy to not speak, but I don’t think we can half-ass examples. The audience won’t understand that these are temporary/only working now. I’d like to advance these into full examples if we put in the work to make them vignettes. On a positive note, this will act as a stress test of the infrastructure (the same sort of thing happened around the bespokefit workshops)

    • JW – For the stuff we show in the talk, I’m OK to say “this is only working now, expect this to break” for some of them. This is where we can use the “Green light/Yellow light” system. Also feel free to assume a maximalist and fully pinned conda env

    • MT – Even in a maximalist env, I’m worried that interactions with Chodera lab packages (eg yank and evaluator) will cause a lot of brittleness

      • JW – Ok, let’s deprioriitze evaluator stuff. We can reassess wheether this becomes low-hanging fruit.

    • MT – How much time to spend?

      • JW – MAybe 1/3rd of time between now and meeting - So ~5 days.

      • MT – It will only make sense to have those be the last 5 days.

    • JW – I’ll prepare vignette slide layouts, and will indicate what you should fill in. But for the simulation ones, if you can get them to a simulation-ready system I’ll do the running and video-ing.

    • MT will work on vignette production for ~5 days, starting closer to the annual meeting time

Trello

 

Issues and PRs

 

Action items

Decisions