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 stereo-dependent SMARTS

DD –

• Protein-Ligand Benchmarks

  • met with folks from OpenFE, obtained consensus around MVP prototype over next few months

  • 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

  • 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

• 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

JW –

• Topology refactor PR ready for review

  • Next will work on nicer interfacing with Openmm

• JMitchell is moving forward with user docs