Possible mis-ordering in carbonyl bond parameters

[bannanc]

Here is an e-mail correspondence considering how parameters are assigned to C=O bonds in protonated carboxylic acids compared to carboxylate groups.

Hi Chris,
Does the “[#6X3:1](~[#8X1])~[#8X1:2]” string represent the C-O bond in carboxylic group? Also is "[#6X3:1]=[#8X1+0,#8X2+1:2]" the corresponding string for the carbonyl bond C=O?
If so, it seems to me that using your paradigm “last wins” the first string should go after the second one, otherwise the parameters of one C-O bond will be overwritten by the parameters for C=O bond, and the carboxylic group will be asymmetric.
Do you agree?
Stan

Hi Stan,
The first string “[#6X3:1](~[#8X1])~[#8X1:2]” os for a carboxylate and the second "[#6X3:1]=[#8X1+0,#8X2+1:2]" is for a carbonyl group or a protonated carbonyl group.
You are right the second would overwrite the carboxylate C=O of the first, and reorodering it like you say would remove that bad behavior. You are probably right... I would like to try out your suggestion and see if it is better everywhere.
Christopher

David and I agreed that this seemed reasonable. Note, this sparked two discussions one for the parameters for carboxylate groups and how we evaluate changes to a SMIRNOFF file. This issue will address the former and the latter will be addressed in its own issue (potentially in the openforcefield repository as that is a test we will want for any SMIRNOFF formatted force field now or in the future).

[bannanc]

I had a memory of us making changes to this section before, but I think it was pull request #11 which addressed the difference is C-O bonds in carboxylic acids and esters.

[davidlmobley]

This also raises a larger question of how we test a change like this. Any thoughts? It seems to me we might ultimately be interested in two levels of tests:
a) Scan a set of molecules and see which molecules it affects and check (manually?) they are those we expect it to affect
b) Repeat our validation work for the affected compounds (e.g. dielectric constants, densities, hydration free energies)

Thoughts?

[bannanc]

@davidlmobley

I agree we need to also discuss that part, I was going to write a separate issue for how to validate changes.

[bannanc]

Here's some details as a part of the carboxylate groups:

Here is the section that is relevant part of the smirnoffishFrcmod.parm99Frosst.txt file:

[#6X3:1](=[#8X1])-[#8X2H0:2]  320.0    1.340   Frosst C_-OS force ff94 CT-OS ; R0 xray. SMIRKS updated by DLM to not match C_-OH
[#6:1]=[#8X1+0,#8X2+1:2]      570.0  1.229   parm99 carbonyl oxygen
[#6X3:1](~[#8X1])~[#8X1:2]      656.0  1.250   parm99 carboxylate C~O

This looks like it is in the correct order, as Stan noted [#6X3:1](~[#8X1])~[#8X1:2] should be after [#6:1]=[#8X1+0,#8X2+1:2] so the two C~O bonds in a carboxylate group will be treated identically.

Here are the corresponding lines in smirnoff99frosst.ffxml:

    <Bond smirks="[#6X3:1](=[#8X1])-[#8X2H0:2]" id="b20" k="640.0" length="1.340"/>
    <Bond smirks="[#6:1]=[#8X1+0,#8X2+1:2]" id="b21" k="1140.0" length="1.229"/>
    <Bond smirks="[#6X3:1](~[#8X1])~[#8X1:2]" id="b22" k="1312.0" length="1.250"/>

@davidlmobley and @cbayly13 could you confirm that is this the order we want things in? I'll run a couple of example molecules, but I think this is what it should be.

[davidlmobley]

I'm slightly confused. So you're saying that the order is already correct? So Stan may have been looking at an outdated file with incorrect order?

[bannanc]

I haven't looked too far back, but yes, I think the order is already correct (and in the order Stan suggested it should be).

[bannanc]

Since I'm looking at this section, I think [#6X3:1](=[#8X1])-[#8X2H0:2] should be [#6X3:1](=[#8])-[#8X2H0:2] or [#6X3:1](=[#8X1+0,#8X2+1])-[#8X2H0:2] so that it would match the corresponding bond if the carbonyl oxygen was protonated. @cbayly13 do you agree?

[bannanc]

I'm trying to find when this was changed:

Order was incorrect when this repository was made (first commit)

It was updated in PR #43 where there were a lot of changes

You can see our full discussion on this same topic in issue #38!

[davidlmobley]

Haha, excellent, so we fixed it before Stan caught it -- basically just an outdated version of the file -- which means all the validation work was done with the correct FF. :)

[bannanc]

yep, I would like to think about the other parameter I brought up:

Since I'm looking at this section, I think [#6X3:1](=[#8X1])-[#8X2H0:2] should be [#6X3:1](=[#8])-[#8X2H0:2] or [#6X3:1](=[#8X1+0,#8X2+1])-[#8X2H0:2] so that it would match the corresponding bond if the carbonyl oxygen was protonated. @cbayly13 do you agree?

It is definitely a corner case, but I think our general rule is to have the earlier SMIRKS patterns as general as they can be.

[bannanc]

I'm going to close this since the ordering was fixed and put my smaller protonated carbonyl question in its own issue.