2021-10-11 Core Developers meeting notes

General updates

• MT – New Psi4 release

• JW – Maybe also a problem with pycpuinfo again with the new release.

IP –

• Working on from_pdb API point. Got it to a “prototype” state.

• • Can read a T4 lysozyme from pdb and turn it into a chemical OFF molecule in ~30 seconds.

  • Very messy code currently, copied directly from notebook.

• JW – I’d be in favor of merging this kinda early, even if there are remaining to-dos.

• We now have an arsenic conda-forge package, thanks to Will Glass and Mike Henry https://anaconda.org/conda-forge/arsenic

CC –

• PI wrangling for LiveCOMS review. One dropped but we can probably replace them with one of DM’s alumni.

• Re: Biopolymer call – Working on librarycharges for rosemary FF. Trying to get polymeterizer running on my machine, then will be looking at different protonation states supported by FF14SB.

• For dipeptide torsiondrives, DDotson will help figure out how to constrain non-driven dihedrals during torsiondrive. Important thing here is to ensure that this information gets propagated correctly through QC stack.

• Out Weds-Fri.

  • MT – Recommend not looking at work during this time, possibly leave someone with your phone # so they can call if there's a “computational chemistry emergency”

MT –

• Put some work into design guide for Interchange - will probably get merged in an early state after some iteration with JMitchell.

• Project scoping meetings

• Worked with JW on current state of topology refactor branch, but didn’t get far. Will try again.

• Worked on cleaning up GROMACS parser, tried to remove dependence on InterMol and bring more functionality/structure in-house. Most functionality is now moved and all the tests are still passing

• Worked on AMBER exporting. Lots of decisions left to make, but I’m working on replicating ParmEd behavior/test cases. I’ll continue working on this in conjunction with topology refactor. Optimistic that we can get protein export working sanely in the near future

  • Packaged the amber FF14SB port - The forcefields from the amber_ff_porting repo seems to be intended for production use, so I’ve made a conda package of those

    • JW – Ah, this may motivate fixing the improper indexing bug in the released file:

  • CC – Some of the DCerutti torsiondrives have D- amino acids instead of L-. This makes them unparameterizeable, since the SMARTS in the FF are stereo-dependent

  • JW – Can we safely remove all @ signs from the protein FF? Will it still work for librarycharges and torsions?

  • CC – Librarycharges should be transferable in a mixed peptide. Torsions will require a study to decide.

  • PB – Could you recap background about librarycharges?

    • CC – If a protein has a mix of L- and D- amino acids, can the same librarycharges be used?

    • PB – This is different from the point about deriving librarycharges using polymeterizer?

    • CC – Yes, this is more of a technical thing

DD –

• Protein-Ligand Benchmarks

  • met with folks from OpenFE, obtained consensus around MVP prototype over next few months. Depending on scope this could require a really large investment from everyone involve, so in the next few months we’ll look toward achieving minimal functionality and letting stakeholders determine what they need/can provide with less uncertainty.

  • development effort from us (me) on this front

• QCArchive

  • pushing for public QCArchive upgrade to fix some ongoing production issues

  • submitted Willa Wang's RESP polarizability dataset; getting consistent errors for cases with imposed field

    • will investigate this week

  • JHorton pointed out an issue with some workers, possibly due to QCEngine/psi4 release. I’ll be following up with this.

  • worked with Pavan to set up prod managers on his UCI HPC3 resources

  • worked on user quickstart for qca-dataset-submission; not quite finished adapting OptimizationDataset submission starting point. Goal is to make it a lot easier for new users to get started in the submission process. Once we get feedback about the first version we can get refine and include instructions for torsiondrives, etc.

• Benchmarking

  • still need to pull together partner retrospective survey; on docket for tomorrow to get this out the door

  • need to follow up on openff-gopt interest

PB --

• Writing the QM theory benchmark paper:

  • https://docs.google.com/document/d/1tMoJY0J2E6uRltvL4gJv7ogW-sdNgm4DB5lcl4LFMNU/edit

  • PB – Feel free to provide additional comments, or let me know if there’s anything more you’d like to see. But this isn’t suitable yet for formal review.

• Running some fitting experiments, will continue working on those this week.

JW –

• TopologyMolecule removal PR ready for review

  • Next will work on nicer interfacing with Openmm

• JMitchell is moving forward with user docs. Seems to be working one week for OpenMM, then one week for OpenFF, so let’s see if we can iterate rapidly this week and get things kinda wrapped up by Thurs/Friday.

• Will be working half time next week (8 AM - noon US Eastern time)

• Will be leading roadmap review at Wednesday’s all-hands meeting.

MT – Toolkit release? I think that, if we don’t get one out before November, then we’ll get busy and won’t make a release until ~February.

• JW – Agree. I’ve put the 0.10.1 toolkit release on the calendar for two weeks from today.