2021-04-16 WBO/Impropers meeting notes

Date

Apr 16, 2021

Participants

  • @Pavan Behara

  • @Simon Boothroyd

  • @Jessica Maat (Deactivated)

  • @David Mobley

  • @Christopher Bayly

  • @Trevor Gokey

Goals

  •  

Discussion topics

Time

Item

Presenter

Notes

Time

Item

Presenter

Notes

 

Impropers

@Jessica Maat (Deactivated)

  • (JM will link slides here)
    JM: I started working on the Openff aniline para, plan is to download the dataset and run geo opt with Openff-1.3.0, and compute their improper angles

  • CB: And the valence angles also for the same structures?

  • JM: You mean summing up the valence angles?

  • CB: Yeah, for checking the pyramidalization

  • DLM: JM and I were talking about this few days back where Nitrogen cases go either planar or tetrahedral not in between.

  • CB: I have the same issue for a long time

  • DLM: So, adding a few more molecules to scan that are squishy would allow us

  • CB: Looking at the minima for different substituents on the spectrum of pyramidalization would inform us

  • JM: I was comparing the improper angle and wbo correlation based off the assigned parameter, analyzed the grid opt nitrogens nad created the scatter plots, the overall thing is that there isn’t any correlation

  • CB: So from the smarts pattern any Nitrogen bound to the sulfur/phosphorus

  • DLM: Let’s make these plots interactive

  • JM: Yeah, I am gonna do that

  • CB: So, the wiberg bond orders are double bonds?

  • JM: They’re the sum of wbos here

  • CB: Aniline set?

  • JM: No, this is an old one from Nitrogen grid optimization with phosphonamides, sulfonamides

  • CB: I would want to see the improper angle is more planar when the nitrogen is more conjugated, and the relevant bondorder is between the nitrogen and the phenyl

  • DLM: How do you know without looking at the molecule the bond order of a molecule that is conjugating?

  • CB: If I can use smarts then nothing Nitrogen/Carbon cannot conjugate, and Sulfur/Phosphorus, anything with #X4 with a hydrogen cannot conjugate

  • DLM: What if there are two conjugated bonds like n case of a Nitrogen

  • CB: In case of two phenyls attached to a nitrogen then it would prefer to be planar

  • DLM: So, we bin the bonds coming out of nitrogen and check how many conjugated bondorders are present

  • CB: Put the avg bondorder from the conjugated bonds and the improper angle, when you stick a nitrogen in between two phenyls I know it must be planar but I don’t know how much the conjugation is and how planar it would be?
    As we look at these relationships with WBO, we have to look at the QM optimized geometries I mean how well the QM WBOs will match to the AM1 WBOs

  • DLM: So, we want to come out with the conjugated bonds from the Nitrogen, in this process we can exclude bonds to Hydrogen and #X4

  • CB: I would guess the wbo of N-H wbo would be a good threshold value beyond which we will see the conjugation effects kick in
    We may place a sigmoid curve and that would weigh the larger wbos higher
    The main purpose of WBOs is to answer the questions on planarity, conjugation, etc., and in this process I would encourage to include impropers, torsions, valence angles, simultaneously.
    Let’s look at what the structures tell us
    The sum of wbos on this plot shows that it can be anywhere from 2.1 to 2.6 and still be pyramidal

  • DLM: Yeah, we can improve this plot by picking the conjugated bonds

  • CB: If we got a phenyl that conjugates the nitrogen, then it is somewhat pyramidal, but if we place another phenyl then we may have two conjugated bonds and averaging it we may get the singly conjugated bond but yet this would be planar, so that would be a misrepresentation

  • DLM: Let’s separate out the singly conjugated and doubly conjugated into bins

  • CB: May be we use a geometric mean instead of normal average

  • TG: One thing I want to point out is that these plots are symmetric around zero

  • DLM: Some molecules have handedness due to chirality and may pucker in a different direction but for the purposes of FF yeah we can fold the graph

  • TG: Even when you average?

  • DLM: Yeah, that should work.

  • CB: All this physical chemistry conjugation, bondorder, etc., when you place a methyl in place of Nitrogen you will see a drastic difference in the chemical environment, and the 1-4 sterics effect would push the p-orbitals and the nitrogen would be pyramidal, when we put that methyl on aniline and the confounding sterics going to drive the nitrogen planar, I would suggest some poking around by placing a methyl substituent when you have time

  • DLM: To avoid the supercomplicated cases, we are looking at the cases where MM is definitively better or bad

  • CB: We are trying to rule out the sterics mixing with our wbo work and looking at NH2 instead of NH(CH3), I agree with the science here, I also hope to look at the other side too since wbo would drop when we have a methyl substituent, valence angles and impropers are not gonna look at the rotation of NH2 so let’s keep looking at the torsions too, and the larger picture is how to capture the pyramidalization in our FF, at the design level we should be prepared to handle the torsions too.

  • DLM: JM will be out on an OE internship at the start of June, I hope we will get some chemistries out of her analysis work before that.

  • SB: Yeah, I agree



Periodicities in torsion terms

@Trevor Gokey

  • TG: Everyone stops at periodicity of six, why?

  • CB: Because we don’t want to overfit

  • TG: I found that using higher periodicities allows certain areas in the torsions potential to be flat, which might help describe “soft” chemistry

  • CB: When we found that the 1-4 interactions won’t fully describe the torsions
    Everybody knows that the 12 potential of LJ is too repulsive and cut it into half I think
    Data driving the optimization of scaling factor is from the condensed phase data and with two parameters we can either do steric clashes well or condensed phase
    So, we need to have a three parameter steric potential that would describe the short range (torsions etc ) as well as long range effects

  • TG: So, if I put the vdw on this plot then both the minima and the barriers should increase in energy

  • CB: Yes

  • CB: SB you were going to reparametrize the scaling actor right?

  • SB: Yeah, the infrastructure is there we need a person to work on that

  • CB: TG, are you looking at specific individual molecules or on a full dataset?

  • TG: I am trying to generally apply it.

  • CB: Did you find one?

  • TG: There are splits, yes. I use an analysis to guess the periodicities, and make new torsions if the guess differs by a significant amount between two groups

  • CB: It would make the preference between gauche and anti conformers clear
    anti are more stable and in order to stabilize it over gauche we need a periodicity of one, and if we find something that would null this hypothesis

 

 

@Simon Boothroyd

  • Plotmol has highlighting atoms feature now

  • 1.3.1alpha is out

  • How much do we use ANI? Should we include that in the benchmarking work?

 

 

@David Mobley

Nicholas Moitesseir papers where they used electronegativity to describe the conjugation effects

  1. https://doi.org/10.1021/acs.jcim.9b00585 (Biaryl)

  2. https://doi.org/10.1021/acs.jcim.0c00440

  3. https://dx.doi.org/ 10.1021/acs.jcim.7b00645

Action items

Decisions