2023-12-13 All-hands meeting notes

Infrastructure

JW

• QC* stack 0.50 rollout

  • QCSubmit has been updated

  • Recharge+BespokeFit+OFFTK are still in the works

    • CC – Details about OFFTK incompatibility?

    • JW – From_qcschema method behavior needs to be audited, likely there are behavior changes that we need to fix/document.

  • qca-dataset-submission is open for business again.

    • We’ve run a test optimization and singlepoint+wavefunction dataset and everything seems to be fine. Feel free to prepare submission PRs, and attend the “QCSubmit discussion” calls on the OpenFF calendar to get them reviewed and submitted (next one is Dec 19 at 1 PM Los Angeles)

• We’re aware of a possible issue with openff-utilities for micromamba users.

  • We’ve observed this as BespokeFit erroring out when it’s unable to determine the ambertools version, on systems with ONLY micromamba (and not conda or mamba).

  • We’re unsure about severity - Has anyone on this call experienced this?

  • A fix is in the pipeline and may be released as early as tomorrow.

• OpenEye asked us to include Xe vdW and LibraryCharge parameters from this paper in Sage, and we’re likely to release openff-2.0.1 and 2.1.1 with these changes shortly.

• PRELIMINARILY We may change how charge assignment is done using OFFTK with the ambertools+rdkit backend in order to improve consistency and quality. This would likely involve using an “ELF1” conformer and specifying additional keywords to antechamber.

  • Background: In running replicate free energy calculations of the same protein-ligand complex multiple times on different hosts, OpenFE is finding that OFFTK’s AM1BCC partial charge assignment can produce different charges under different circumstances, leading to significant differences in predicted free energies. In their ongoing study, this appears to be due to

    • RDKit generating different conformers

    • antechamber optimizing the same geometry to different minimum energy structures

  • in different circumstances, like

    • different system loads at the time of calling these programs

    • different system architectures

    • different OSes/system libraries

  • The OpenFF lead team discussed this and we are very likely to approve+implement OpenFE’s specific recommendations once they present them to us.

  • CC – Agree this is a good change. Path forward should be changing the SMIRNOFF spec and releasing new versions of the FFs with the new behavior specified.

    • JW – Interesting, we hadn’t considered re-releasing FFs.

    • DM – Since FFs were trained using OE backend, this might not necessitate re-releases of FFs.

    • CC – Since this changes the behavior of the ToolkitAM1BCC tag, it could be argued that this isn’t just an implementation detail, and should be reflected in the spec/FF contents.

    • DM – …

    • CC – Would involve changing ToolkitAM1BCC version=0.3 to 0.4 and update the spec

    • (looked at spec)

    • DM – Agree with Chapin, we should update the SMIRNOFF spec to specify ELF conformer selection with RDKit/AT backend.

    • DM – We should email OE, Cresset, and other people maintaining SMIRNOFF parsers and warn them that the tag change is very likely coming.

    • JW will contact downstream parser maintainers

• PRELIMINARILY Expect another series of OpenFF “follow up” workshops in Q1 of 2024. Dates+subjects TBD.

OMSF

ZB

• OMSF will need your participation in some future interviews to satisfy POSE requirements

  • ZB – NSF mandates a series of interviews to determine the nature of OMSF ecosystem. Logistically this will be 30 minutes of your time, with myself and Mike Henry, to determine what you envision from an open source ecosystem. Likely to happen post-new years.

  • JW – Just you and MHenry, or will there be NSF people on the call?

  • ZB – Just me and/or MHenry.

  • DM – Part of the requirements for POSE is to have representatives POSE projects meet with people in their ecosystem to

  • JW – What’s the final deliverable from your interviews?

  • ZB – NSF iCorps (icore?) is tailored toward startups and new products, in the interest of helping teach how to go about finding new customers/discovery. Partof this is the “lean startup” method, where they want to run hundreds of interviews to collect info. iCore wants to determine whether some aspects of their lean startup process are applicable to/can me made applicable to open source. So we’re conducting interviews on behalf of the NSF to help them in this process. We’re figuring out details of how exactly we’ll conduct this and seem to have some latitude in that regard. We’ll have more info when we start scheduling meetings in 2024.

  • DM (chat) – My view of this is something like NSF wanting to help entities (like OMSF) figure out how to become sustainable and build their ecosystems.

  • DM (chat) – In our case “sustainable” probably means “bring on more projects and partners and build software that even more people want to use”. So you can think of some of it as like … “who is your market?” stuff probably.

XtalPi/XFF/Pfizer

DM

1. They may convert XFF into our formats to allow easier testing/comparison. (XFF would still be distributed by them under another license but they can do that.)

2. They want to contribute to improving FEP benchmarking datasets, helping to pick cases that are sensitive to force field quality rather than errors being dominated by other sources

3. They would like to be involved more in data sharing relating to QC data, potentially contributing their data and/or participating in some kind of data sharing consortium if one arises

4. (Their slides)

Shouts-out

DM

• JW – Pavan gave a great presentation at PsiCon last Friday!

  • ZB – Please let me know if those recordings become available and I’ll post on socail media!

Invited speakers

TBA

Q&A

DM

DM – OMSF has positions open, please ask me or ZB if you have questions