Ring parameters | Lexie | Slides will be uploaded CBy – which of the “specific” angles are exocyclic? Those are the critical ones, much more so than the internal angles – e.g. there’s not much an angle can be in a 3-membered ring other than 60° (on 4-membered rings slide) CBy – I have a couple low-hanging fruit recommendations on these previous results you’ve presented. Can we just change the ordering so that a completely generic wildcard three-membered ring is 60 degrees period? It shouldn’t matter if it’s a C, N, O, in a 3-membered ring, should be 60 degrees. We shouldn’t need specific parameters in the ring for different elements. Now, with the 3-membered ring, what if we have two connected 3-membered rings – we can specify ring closure like [*:1]1-[*:2]-[*:3]1 (i.e. a3). Similar example for 4-membered rings. AMI – (on whether we may need multiple parameters) e.g. a 4-membered ring with O in the middle of the angle has a much higher force constant, but that may not matter CBy – doesn’t make sense to split out; we could probably group them together and have one good force constant.
DLM (in chat): I have a question about “four membered ring is around 90”; this is a function of pucker, right? So if the rings are more puckered we might see a larger angle…? Would be interested to know how flat those different rings are. CBy – this should likely fall onto torsions to enforce pucker DLM: I don’t know I’d want to enforce equilibrium angles to be equal just because the torsions would cover it CBy – since AMI is pointing out dramatic problems, we would likely get most bang for buck with this (AMI’s) approach first DLM: agree, this should be considered as a possible future approach, we should do the easy stuff first and then go from there
(on 5r angles – MSM vs Sage) (on 5r angles slide) AMI: raised parameter confusion of a18a: central atom is in a non-6-membered ring CBy (in chat): valence parameters of key interest: ringAtom-ringAtom-notRingAtom CBy: are these distributions contained in the ones you’ve shown us already? They should be showing quite marked deviations AMI: I don’t think I saw that in the distributions, but possibly the distributions are very broad so I didn’t see it CBy: I think this would be very important for 3, 4, possibly 5-membered rings. CBy: the force constant can be quite correlated to the angle, especially between a valence angle movement and bond stretching movement. So it may not matter so much for our ffs. But what does really matter are the equilibrium values for the exocyclic angles.
BSwope: found that relative energy of ring conformers was pretty messed up and apparently this is seen in a lot of cases, and possibly because force constants are too stiff and avoid interconversion
|
Parameter splitting | Brent | CBy: factor of 2 in objective function is concerning DLM: I would want to ask Lee-Ping about that change in objective function CBy: should we make the force constant dependent on the equilibrium value? Is that easy science? Although I care much more about equilibrium values than force constants and that’s where we would want to split parameters DLM: suggests assigning unique values for each molecule and seeing performance difference to a general force field Trevor (in chat): If I was debugging the objective, I would determine if it was optgeo or torsion for a single target (i.e. look at the same target per version and see what the delta is). If it is optgeo, then I'd double check the denominators, and if it is torsion, I would double check the attentuation params LW: that linear data you showed, Brent – was that MSM or espaloma data? Did espaloma give the same linear dependence?
|
