2021-08-11 All-hands meeting notes

General Updates

@Jeffrey Wagner

5 min

• We’ve had sage RC1 and RC2. RC2 has behaved well and will become actual Sage release. Final Sage has green light, will go through technical review and be released in coming days.

• Interchange is seeking early users and testers – Allows post-assignment parameter modification, export to OpenMM and GROMACS

• OpenFF toolkit 0.10.0 released. Expect slight differences in energies.

• OpenFF-supported industry benchmarking is wrapping up. Janssen personnel are pushing to also include OPLS and Sage but this is beyond scope so they’ll be driving it.

Follow-up workshop update

@Jeffrey Wagner

5 min

• September 1 and 8 at 7 AM Pacific / 8 AM Mountain / 9 AM Centra / 10 AM Eastern / 3 PM UK / 4 PM CET.

• First will be on Benchmarking, led by Lorenzo. Second will be on Bespoke Fitting, led by Josh Horton.

  • May include interactive parts, TBD closer to then. Sessions will be recorded and materials made available.

AM1 studies update

@Connor Davel

25 min

• DM – “MD constrained” should be “MM constrained”

• DM – Have you thought about a procedure where we’d do an unconstrained minimization, except in cases where a constrained one is needed?

  • CD – Yes. The rest of the decision will be based on timing/profiling tests

• DM – Better reference than openeye? When we looked into this initially, OpenEye people thought that unconstrained would be better when there isn’t a connectivity change. So could we reference this against electrostatic potential?

  • CD – We’ve talked about reproducing other QM simulations. This is a problem because, if an AM1 min has a connectivity change, a QM min would also have that connectivity change. We did think about using solvation free energies as a reference, but this is tricky.

  • DM – That makes sense. But I’m thinking that, for non-rearrangement cases, maybe OpenEye isn’t the best.

  • OM – It’s really ahrd to define a reference, especially because AM1 will be baked into other parameters.

  • DM – I think the real “bake-in” will be in the vdW parameters. Maybe the thing to ask is “if I do an optimization with an OpenFF typical level of QM, then I fit RESP charges, how well do those correlate with different AM1 energies (with some weight also toward considering keeping HFE accurate)?” Generally I’m nervous about using OE as a standard

  • OM – Agree. We should connect up with Chris

  • DM – We have a few issues we’re resolving ehre:

    • The charges shouldn’t depend on the backend, which this is solving

    • The charges should be accurate

  • DM – Also, I’m nervous about coverage DIFFERENCES between our FFs and the AM1 parameter coverage. Because then the set of molecules that we can handle is the SUBSET of the AM1BCC coverage region that intersects with the openFF forcefield coverage.

    • JW – That’s a really good point.

  • JW –

• DM – When I worked with CBayly, we thought about molecules where there’s no rearrangement vs. molecules where there ARE rearrangements, and whether OPenEye’s “constraints in all cases” philosphy is makign those WORSE?

  • CD – We can see this on the RMSE scatter plots. I’m working on making these using plotmol so we can visualize the outliers.

• DM – I’m still interested in finding a better reference method that will let us determine which constraint method is overkill

• CC – I think MSchauperl’s RESP2 approach could be a good target.

• CC – Also, I’m curious, when you do the MM constrained minimizations, I wonder how those results would change with a solvent model

  • JW – That’s a really good idea

• OM – I do wonder how many of these rearrangements are “vaccuum exclusive”?

• DM – Could we set up a electrostatics meeting with Simon, Bayly, and Schauperl (if we can get him)?