2025-06-17 FF Fitting Meeting

Torsions – Mark Mackey

• LW: (type 1 case):

  • Did you re-fit from the manual adjustment?

    • MM: no

JW: do you think the Type 4 interacts with the Type 1 issue as well? Could the cause be the same?

• MM: I think there are a few issues. The wildcard is causing H-C-C-N and C-C-C-N to match the same, but I think they’d probably look quite different (and probably to O-C-C-N too). We want to look into splitting out this torsion.

• JW: Do we think there’s a chemical perception issue with the N? Confusion over whether it should be pyramidal or planar?

• MM – Possible, but I think the answer more generally is the *-C4-C3-* torsion is causing issues. Might be resolved by making more specific to hydrogen/non-hyrogen terminal atoms.

JR: the real molecule has amide hanging off an aromatic ring – do you have examples where it’s not aromatic?

• MM – we see similar issues for any thiol/amide, doesn’t matter whether neighbor is aromatic

• JR – and this is after MM optimization? Which wants to be in plane vs. out of plane?

• MM – QM wants to be flat, MM doesn’t.

DM – Would be interesting to know whether Paul Lebute’s extended huckel work could help here.

• MM – Maybe

BS – My chemistry intuition for the type 5 example is completely failing here - Why is the placement of the sp3 carbon so impactful?

• CB: this is an anomeric effect, although I’m handwaving here. The phenolic O lone pair is interacting with the double bond. I would imagine in-plane O to be sp2-ish, but out-of-plane would be sp3-ish, so I’m not sure here, but otherwise I can’t come up with much explanation. What’s the energy difference?

• MM: 2kcal/mol between min and max

• CB: that’s about the size of an anomeric effect. We could build this into the torsion with SMARTS. But how general is it? How would it affect it if you made it an aromatic ring by replacing a CH2 with N?

• MM – The torsion profile is very sensitive to the specific heterocycle.

• CB – Thinking about kinda arrow-pushing here… Not sure what to conclude. This seems like a difficult problem. The cyclopentadiene is one of the most extreme cases of this anomeric effect because of the effect of the double bonds.

• CB – This is where I’d think WBOs would make a big difference, by identifying that the cyclopentadiene isn’t aromatic, but is squarely double and single bonded. …

• CB – If WBO infrastructure is still working, this could be a good chance to test that out.

LW – I never presented these results, but I ran a very naive initial refit to the new data, and also observed very little benefit. I’m guessing you got the same issues and have done a better job of diagnosing them. But feel free to bring more datasets you’d like run, we’d love to work together more.

• MM – Great, will do. This is also fairly early work, and I’ll be in touch if we have ideas for further datasets.

Ash Sage updates – Lily

(slide 5) DM – Make sure that we don’t let that one H actually get a 0 vdw term, that causes technical issues.

(slide 17) MM – I’m surprised that the specific carboxylate angle term doesn’t seem have have ANY mols that are correct. You’d think we could just add 5 degrees to that angle and make it right.

• LW (hidden slide 21) – It looks like the angles are trying to make the geometry right, but other factors (nonbonded?) are pulling them out.

(slide 14) JC –