2023-10-11 All-hands meeting notes

Updates

JE (or DM)

• Save the date for an OMSF meeting May 16th and 17th in Leiden, NL! We are planning it to immediately follow the Alchemistry Workshop, so if you are planning to attend that meeting, consider extending your trip by two days. Details to follow!

• OpenFF Toolkit 0.14.4 is out, with NAGL in the public API.

• Introduce Hannah Turney

• Infrastructure update (slides)

  • JC – Plans for LJPME?

    • DM – … (see recording ~32 minutes). It’s interesting but we don’t have anyone to put on that.

• Science update (recorded)

  • JC – What integrators / Barostats were used for the OpenMM vs Gromacs?

    • CC – I’m using Langevin middle integrator in OpenMM, GROMACS is using canonical velicoty rescaling from bussy for temperature and pressure

    • JC -0- Those might gitve very different ensembles. You might just try comparing densities and average enthalpies.

    • CC – What was solution? Stochastic integrator in gromacs?

    • JC – We’d need to work with MS to idenify that. The thing with large timetsteps is that you can get timestep-dependeint perturbations. So like BSwope had to extrapolate results to “0 timestep” to get good results.

    • CC – Gotcha - I’ll follow up with MShirts.

  • JC – Chapin: Did we end up excluding the idea of using the QM energies directly without constrained optimization in fitting, which would accelerate fitting experiments? Just curious if strong evidence emerged to discount that approach (which has worked well for espaloma)

    • CC – We decidedf that protein wasn’t the right dataset to test that. We should do this on small mols first, but I’m not sure whether we have a person working on this.

  • JC – And I'd love to learn more about the torsion issue Brent is looking into

    • BW – I’m working on a deeper dive/want to update this plot, but I’m having trouble reproducing the Sage fit.

    • JC – More details on the underlying issue?

    • DM – CBayly got a little too clever when making the initial torsion patterns.

    •  

Recognitions

JW

• successful OpenFF training workshop at CCP-BioSim. Material is available here: GitHub - openforcefield/ccpbiosim-2023  

• DM – Joe Greener gave a good talk on differentiable FFs this morning. Lots of promising areas of possible overlap.

  • DC – Yeah, I think this one could be a good ongoing connection. The Riniker talk was a good one-off, but we have an opportunity to engage with Joe since he’s a young PI.

  • JC – CIacovella and I would love to be inlcuded in followups.

Q&A