2022-05-18 Thompson/Wagner Check-in meeting notes

General updates

• JW – Annual meeting June 28 - I’ll probably hit you up for slides

  • MT – Sounds good. I have scipy two weeks later so I’ll reuse the slides.

• JW – Bespoke release - Expecting a little drama/user support. We’ll see.

  • MT – I wouldn’t stress too much. If there’s something seriously wrong we can battle-harden it and do another soft launch in a few months.

• MT – What’s the future of bespokefit? Do we just keep it running in its current form, or will this become central to fitting team work?

  • JW – I don’t know. I think sage was made largely using an early version of bespokefit (or something like that). So I think that the fitting team may continue to use it heavily.

PR/Issue clearance

• Bespokefit single file installers wrapup?

• Are there valid cases in which a molecule will have multiple charge handlers applied to it?

  • JW – No

  • If charge_from_molecules applies anywhere, the molecule is done since it can only apply to the WHOLE molecule

    • JW – Yes

  • I can’t think of any useful cases in which library charges are applied to a part of a molecule but not the whole (ionic molecules, biopolymers, LibraryChargeType.from_molecule

    • JW – Yes. LibraryCharges only apply to a molecule if they cover EVERY atom

  • I can only think of cases in which ChargeIncrementModelHandler is applied to an entire molecule

    • JW – Yes

  • If AM1BCC is used, does it even make sense for the molecule to be a patchwork of the three charge methods?

    • JW – No

0.11 release planning

• PDB reading

  • Speed

  • Input diversity

  • Better errors (loudly yell about what we require from inptus)

  • Better docs (loudly yell about what we require from inptus)

  • from_rdkit/openeye (assigned stereo and aromaticity?)

• Examples

  • Protein FF finalization (should people be using the ff14sb port)?

  • “Happy path” for protein+ligand setup =

protein_interchange = ff14sb.create_interchange(protein)

everything_else_interchange = sage.create_interchange(ligand_and_waters_and_ions)

final_interchange = protein_interchange + everything_else_interchange

omm_sys = final_interchange.to_openmm()

Notably, the recommendation on release day will NOT involve importers from openmm/amber/gromacs.

• Migration guide

  • Units (done)

    • We can load OpenMM.units, but NOT simtk.units

  • Topology stuff/API breaks

• Downstream package releases

  • ForceBalance

  • BespokeFit

  • QC*

• API cleanup

  • default hierarchy schemes

• Topology positions/conformers

  • If all molecules have at least one conformer, Interchange will assume that that should be their positions. But if ANY molecule doesn’t have a conformer, then the Interchange isn’t built with positions

  • We will NOT add Topology.positions to the API (at least not now)