2023-02-09 Protein FF meeting note

Amide torsion parameter fit

Chapin Cavender

• SLIDES HERE

• DM (slide 3): this slide shows results on training data – what about test data?

  • CC: these fits are still in progress, and also generating new QC data

  • DM: but these preliminary fits look promising

  • CC: I agree

Grant Timing

Grant is submitted March 6th. What do we think will be done in terms of protein testing? what will be done in terms of posting of preperint / submission of paper to LiveCoMS?

• MS: March 6th is date for NIH, not internal deadlines

• MS: the question is: what can we realistically communicate as preliminary results, 3 weeks from now?

• MS: desired – preprint of the protein collaborative discussions, as that’s one of the aims

  • CC: It’s feasible to have a preprint up by March 6th. We need to draft the introductory text for our current draft. We also need a summary section for the data sections.

    • MS: can we put this preprint in Just-in-time information and update a month after?

    • DM: Probably yes, just before the council meeting, but I’ve never done that. It might make sense in this case.

    • DM: what we want for proof-of-concept in the proposal is slides etc. The proposal is the only thing we can guarantee people will read. We also need the draft of the preprint up. We can use CC’s greatest hits slides for the proposal. We can update in 2 weeks if there’s more.

    • CC: we have the validation QC sets and show that we perform better than ff14sb. In 2 weeks we can maybe have some preliminary NMR benchmark data. Do we want thoroughly sampled short peptides or poorly sampled large proteins?

    • DM: short peptides

    • CC: will do, can have a short draft up.

    • DM: MG may have done this before

    • MS: Priorities – let’s preference completed FF stuff over in-progress community preprint

    • CC: Probably ok to put up a preprint – barriers are that an author may not have a chance to look at it fully. Will prioritise NMR benchmarks on shorter peptides, then work on preprint review.

    • MS: Should be okay to say that we have a benchmark up and running as a “done” aim.

    • CC: how to share data/documents?

    • MS: any method should be fine.

  • Second goal: demonstration of protein FF. We won’t get it fully validated and tested, but what is the likelihood of submitting a preprint? How far along will we be?

  • MS: can we parallelise any of this process with additional help, e.g. a grad student, such as running benchmarks?

    • CC: what I’m doing now for proteins should work for nucleic acids, some manual preparation/modification might be necessary

    • MS/CC: maybe we need to have a discussion on what needs to happen with nucleic acids

    • CC: can brainstorm who to reach out wrt expertise on nucleic acids

    • CC: one idea for additional directions: could test different methods for calculating NMR observables – e.g. could Andrew White’s ML method for chemical shifts work better?

single molecule fit update

Trevor Gokey

• (shows torsion profiles)

• CC: I recall you previously saying that the automatic torsion parameter generation would generate a lot of torsions with different periodicities to get out of minima and then pruned them later

  • TG: I think there was something in the QC TDs it couldn’t fix, so it didn’t have a chance to prune things out

  • TG: started at 0.001 as K value and used simulated annealing

  • CC: seems like having a single periodicity of 3 is enough to model the TorsionDrives

  • TG: yes – the TDs here might be simpler – but not sure why the TDs here aren’t fit that well