2021-03-05 WBO/Impropers meeting notes
Participants
@Christopher Bayly
@David Mobley
@Jessica Maat (Deactivated)
@Simon Boothroyd
@Trevor Gokey
@Pavan Behara
Discussion
DLM: PB or SB have any updates?
PB/SB: Nope.
DLM: Okay, I want to put the impropers fitting back on the roadmap for FF releases post-Sage, I think we were aiming to include it for Rosemary or Thyme.
CB: Looking at these structures from ff-release meeting
I would like to add that amides, carbamates, and other hetero groups would make it flat but Urea here makes the nitrogen at the 8th position in left molecule pyramidal. Being attached to the electron withdrawing pyridine it might be sp2.6 where as the right one would be sp2.8.
When looking at impropers Urea-nitrogens should be separated.
DLM: Yeah, the angles are too stiff too and the proton might be pushed differently.
CB: So, we have four interactions here
wiberg bondorder related
improper pyramidal of N
non-bonded steric interactions
valence angles
DLM, do you suggest removing the stiff improper?
DLM: Yeah, in a way. That has never been fitted and a value of 10 is being used based on Amber FF, if we remove that or fit to a new value then it would make it easier for wbo stuff. Also, we were fitting only the TIG parameters and not all valence parameters.
CB: Hm, readjustment of angles and a lot better world with LJ refit would reduce some noise then. I remember the amber setting coming from Benzene, it was somehow generalized for all.
DLM: Yeah, and improper fitting is not a priority for Sage.
SB: Did you do any improper fits before?
DLM: JM was working on it before.
JM: Yeah, we generated 1D and 2D scans with the impropers and valence angles changed simultaneously. Comparing the MM energies we saw improvements.
DLM: May be JM should slowly transition back to doing improper fitting, that would be useful in informing the wbo work too since we see these cases here with Nitrogen.
JM: To add more on this, I made a grid opt dataset of 1D scans that can point us to chemistries where impropers make a difference.
DLM: Great! If possible can you do a quick statistical analysis and also summarize the work done so far by next week’s meeting.
SB: That would be great to know the work done so far, we can plan accordingly. Also, this will reduce some noise in our wbo work.
DLM: Yeah, it may help torsions.
CB: So, we now have a panoply of torsions, valence angles and impropers
DLM: Yeah, JM did the impropers work changing the valence angles and impropers simultaneously but we were not using forcebalance at that time, and our set of molecules was very small.
CB: If we are doing these deflections we need to bring back the vibrational frequencies too, since they represent the changes near the minima very well locally. We can do one molecule and understand all these interactions in a Kollman way of working on simple systems and generalizing.
If it is easy to introduce torsions into the work JM was already doing with angles and impropers that would be a great start.
DLM: JM, you’re looking at bespoke torsions now right? What if you slowly transition back to doing impropers so that you can finish the paper?
JM: Discussion sounds good but what would be my action item here then?
DLM: We came up with few impropers when we were working on it, and one or two of them can generalize well, IIRC one of them is a nitrogen environment with sulphur nearby. If we can get one new improper that would add value.
CB: Do we have a roadmap for Sage and how does this impacts valence refit?
SB: Release a beta version by early May and provided we see significant improvement we will do a general release by June.
CB: LJ refit and 1-4 interactions do overlap with impropers work. Is there time to include JM’s work?
SB: I don’t want to put undue pressure on JM since we only have 2-3 months, and also it was not in our initial plan. If we identify one new improper that would be great but again no pressure on JM.
DLM: I think we already have one parameter that can be generalized. JM you have a dataset right, if you can summarize next week then we can plan on how to take it ahead.
CB: The greatest pain point would be the slightly out-of-plane cases, experimenting with the above structures without the wiberg bondorder terms we can isolate the non-wbo interactions. Also, we have to check on how vibrational frequencies can help here.
SB: Yeah, we should sync up with Hyesu on this, it would be worth it if we get good yields.
DLM: We have the data, so we can look into the status of work and decide next week.
Action items