problem adding a residue to AMOEBA xml

[Daniel Konstantinovsky]

Hi OpenMM,

I have a protein with an amide (NH2) capping group that I want to simulate with AMOEBA. Unfortunately, the OpenMM AMOEBA xml lacks an NHE entry. I added one like this:

<Residue name="NHE">
<Atom name="N" type="7"/>
<Atom name="HN1" type="152"/>
<Atom name="HN2" type="152"/>
<Bond from="0" to="1"/>
<Bond from="0" to="2"/>
<ExternalBond from="0"/>
</Residue>

but I get the error "ValueError: Atom N of NHE 8 was not assigned."

I've made sure that the spacing is all like the other residues. I've tried putting the entry in different places in the Residues section, but none of it works. I checked that for the other residues, the Residue entry is the only one that tells the FF about the residue. Everything else (bonds angles dihedrals) is based on atom types. What am I doing wrong?

Thank you!
Dan

[Peter Eastman]

That error message is a bit too vague! It's coming from the AmoebaMultipoleForce, which isn't able to assign multipole parameters to that atom. Parameters are assigned based on the type of that atom (the "type" attribute in the <Multipole> tag) and the types of up to three other atoms that are bonded to it (the "kx", "ky", and "kz" attributes). There's no existing tag that matches it. You need to somehow come up with multipole parameters for that atom and add a <Multipole> tag to specify them.

[Daniel Konstantinovsky]

Ok I will try that. In general, would you recommend doing this or just using the NME cap instead? It seems hacky...What do you think?

[Peter Eastman]

It depends on how you plan to generate parameters for the new residue. The standard AMOEBA force field only provides parameters for a fixed set of residues. If you want to simulate anything else, you really should go through a proper fitting procedure to determine a complete set of consistent parameters for it.

And of course, I don't know why you want to use NHE instead of NME. If you have a good reason for preferring it, then it may be worth the effort of parametrizing a new residue. If not, then it's probably better to stick with the one provided by the standard force field.

[Daniel Konstantinovsky]

NME vs NHE shouldn't be a huge deal in terms of my results (nonlinear spectra), although my experimental collaborators use NHE so it would be optimal to be consistent. For now, I took the parameters from the two amide hydrogens of ASN for the NHE hydrogens and NME's nitrogen for the NHE nitrogen based on chemical similarity. Does that strike you as a bad idea or okay? I really want to avoid the enormous headache of parameterizing NHE for AMOEBA if at all possible...Please let me know what you think. Btw, I really appreciate your help over the years !

[Peter Eastman]

You can think about it a couple of different ways. In terms of your results, it probably doesn't make much difference. Assuming you do it carefully and validate your results, you can probably hack something together for NHE, and it will work well enough for your purpose. At least as well as NME. Just make sure you don't do anything egregious, like getting the total charge wrong.

Or you can think about how you'll write it up in the paper. "Because AMOEBA does not provide parameters for NHE, we terminated the chain with NME instead." No one is likely to object to that. "Because AMOEBA does not provide parameters for NHE, and properly parametrizing it seemed like too much work, we hacked something together by hand by copying parameters from other residues that seemed similar." You might get away with it, but then again, you might get a reviewer who reads it and says, "You did WHAT???" Using NME is an easier choice to defend.

[Daniel Konstantinovsky]

Yeah, reviewers wouldn't like that hahahahahahaha. I'll try NME.