Is it possible to treat nitrogens equivalently in, for example, [NH2+]=C(N)C ?

[therealchrisneale]

When I parameterize 1-Iminoethanaminium ("[NH2+]=C(N)C") with openff using the code below, I find that the two C-N bonds differ (as do charges). Is there a way to get a resonance-based parameterization, for example like most protein force fields treat the side chain oxygen atoms of unprotonated glutamic acid identically? This seems similar to https://github.com/orgs/openforcefield/discussions/2 ("Aromaticity of Imidizoliums"). Although I'd be interested to understand if there is a specific solution for this simple molecule, my interest is the general behavior that is exemplified by this toy example.

from openff.toolkit import ForceField, Molecule
import openmm
chemical = Molecule.from_smiles("[NH2+]=C(N)C")
chemical.generate_conformers()  
topology = chemical.to_topology()
forcefield = ForceField("openff_unconstrained-2.2.1.offxml")
system = forcefield.create_openmm_system(topology=topology)
with open('system.xml', 'w') as f:
    f.write(openmm.XmlSerializer.serialize(system))

The two C-N bonds are not equivalent:

                               <Bond d=".13094321625219998" k="422616.24377131934" p1="0" p2="1"/>
                               <Bond d=".13895946609589999" k="275382.57341367257" p1="1" p2="2"/>

Given this atom order:

        <Particles>
                <Particle mass="14.00672"/>
                <Particle mass="12.01078"/>
                <Particle mass="14.00672"/>
                <Particle mass="12.01078"/>
                <Particle mass="1.007947"/>
                <Particle mass="1.007947"/>
                <Particle mass="1.007947"/>
                <Particle mass="1.007947"/>
                <Particle mass="1.007947"/>
                <Particle mass="1.007947"/>
                <Particle mass="1.007947"/>
        </Particles>

Charges on the two nitrogens also differ, though I have other ways to fix that:

                        <Particles>
                                <Particle eps=".7016212989374017" q="-.46641818181818184" sig=".3206876023663901"/>
                                <Particle eps=".3635030558377792" q=".45898181818181816" sig=".34806468869450646"/>
                                <Particle eps=".7016212989374017" q="-.4505181818181818" sig=".3206876023663901"/>
                                <Particle eps=".45538911611061844" q="-.20021818181818182" sig=".337953176162662"/>
                                <Particle eps=".058955968900150965" q=".32988181818181816" sig=".11034276772973167"/>
                                <Particle eps=".058955968900150965" q=".32988181818181816" sig=".11034276772973167"/>
                                <Particle eps=".058955968900150965" q=".32988181818181816" sig=".11034276772973167"/>
                                <Particle eps=".058955968900150965" q=".32988181818181816" sig=".11034276772973167"/>
                                <Particle eps=".06602135607582665" q=".11288181818181818" sig=".26445434132681245"/>
                                <Particle eps=".06602135607582665" q=".11288181818181818" sig=".26445434132681245"/>
                                <Particle eps=".06602135607582665" q=".11288181818181818" sig=".26445434132681245"/>
                        </Particles>

Thank you,
Chris.

system.xml.txt

[therealchrisneale]

In a broader context, this question also relates to: openforcefield/openff-forcefields#125

[lilyminium]

Hi Chris, thanks for raising this issue. It is indeed very linked to #2 and David's comment is probably still the best summary of the situation. Right now the best immediate or short-term solutions we would have are clever SMARTS pattern matching, which is fairly case-specific and ultimately limited. David and Trevor mentioned some solutions to the problem that would be somewhat general solutions, but both would take a bit of scientific effort so are probably not coming soon.

I'm interested that the nitrogens had different charges (I presume using OpenEye?). Using OpenEye v2024.2.1, I seem to get symmetric charges across the molecule when I try to reproduce. Also FYI, NAGL also treats the molecule symmetrically.

[therealchrisneale]

Thank you very much for your comments Lily. The charges I reported were from the base openfe conda env, without OpenEye. OE does give me symmetric charges, which was my "I have other ways to fix that". We're still stuck with the asymmetric bond and angle terms (and presumably dihedral, though I did not check). Thanks also for the suggestions on ways we could fix this and the heads up that a general openFF fix is not likely in the near term.