Dear Jefta,
you are right about your approaches:
Either you extend an (existing) amino acid residue or you define a new one.
There are, however, further possibilities like:
- since your ligands are connected to cysteine, you could simply use the
CYX residue and add the ligand part as additional residue; by
constraining the atom charges during RESP derivation to this moiety, you
could create a ligand library that could easily used as extension to
existing CYX.
- you could copy the CYX residue definition and create a single
constrained atom charge set, which is derived from several/all ligand
extensions; then you still had a ligand library with one special
cysteine residue.
- you could create a modified CYX residue with ff19SB parameters for the
protein part, and gaff2 parameters for the ligand part (e.g. mixed
parameters), but with atomic charges polarized over the complete
residue, maybe even using different conformations for charge derivation.
With that, you'd create a modified cystein residue for each of your ligands.
The junction needed additional parameters, of course.
The best way for ligand parameterization depends on your actual
intention, like specificity, transferability, consistency etc. but also
on your set of ligand systems, e.g. how different extensions polarize
the core cysteine residue charges.
And, of course, you should always try to validate the new parameters,
i.e. whether they "work" for your needs.
Maybe that helps.
Best,
Anselm
PD Dr.rer.nat.
Bioinformatik | NHR.FAU
Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU)
Germany
Am 19.02.26 um 13:57 schrieb Jefta Wucherpfennig via AMBER:
> Dear Amber community,
>
>
> for my PhD project I am working on covalent ligand MD simulations.
> Currently I am trying to figure out, what the best way of modeling such
> a system is. I believe there are two general approaches:
>
>
> 1. use a protein force field (e.g. ff19SB) for the protein, GAFF2 for
> the ligand and derive custom parameters for the junction (an frcmod file
> with parameters for mixed atom types). Create the bond in leap.
>
>
> 2. define a modified amino acid in a prepi/library file and patch it
> into the backbone of the protein. Thereby the whole adduct would be
> described by GAFF2 and one would need to define parameters at the
> junction to the protein backbone.
>
>
> I want to use QM+RESP to derive the single point charges for the ligand.
> I would argue that the second approach might be more accurate for charge
> derivation, and also less sensitive to custom parameters, as the
> backbone is more rigid than the side chain. I would validate that GAFF2
> can produce accurate backbone terms, of course.
>
>
> Given a prepi file of the capped (NME/ACE) amino acid-ligand adduct, I
> can manually change the backbone atoms to use ff19SB atom types and then
> use parmchk2 to fill missing parameters for the mixed atom types with
> GAFF2 analogies. Thereby I have created an frcmod file containing
> junction parameters. I can also define a new atom type in chemical
> analogy to CX for the alpha carbon of this adduct, in order to prevent
> the frcmod from overwriting standard parameters for the whole protein.
>
>
> The adducts I want to simulate are cysteines covalently bound to
> acrylamides, like pdb code 6oim for example. Does this approach seem
> sensible to you, or am I missing something obvious? I don’t just want
> working parameters but really investigate parametrisation for covalent
> ligands. Any experiences, references, or example workflows would be
> extremely helpful.
>
>
> Thank you so much!
>
> Jefta Wucherpfennig – Universität Würzburg
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Feb 26 2026 - 05:30:02 PST
