Re: [AMBER] Non-standard residue parameterization questions

From: Dr. Anselm Horn <>
Date: Wed, 29 Dec 2021 10:57:17 +0100

Merry Christmas

Dear Matthew,

the reason for the discrepancies in the different charge sets certainly
results from the fact that different input structures were used for the
QM-ESP calculations, i.e. your structures are (a bit) different from
those used for the original charge derivation. Since you know that the
QM-ESP is (very) sensitive to the molecular structure, you'd need the
exact same two molecular structures in order to obtain the same RESP

In your case, you created two dipeptides via leap and adapted the
backbone angles accordingly, and these structures you directly submit to
ESP calculation without any prior geometry optimization: your structures
are just based on the leap-template.

Thus, a different result for RESP charges might be expected, if one uses
structures different from those used for the original derivation.

Having said this, I'd like to refer to prior discussions on this list
about that topic, where it had debated how to re-do the original

For your own work, I think that the basic workflow seems ok to me, with
the exception, however, that you should have a closer look at your
structures: You might want to relax them via QM geometry optimizations,
check for true minima, and ensure via visual inspection that there's no
self-interaction between the sidechain an the mainchain atoms, which
would disturb your ESP. For very large and uncommon sidechains, you may
want to include not only two conformers but more.

Since there are many publications with additional parameterizations, I
recommend reading the respective methods sections and supplement files.
(Unfortunately, parameter derivation is not always described in such a
detail that would allow simple repetition.)

Hope that helps


Bioinformatik | NHR.FAU
Friedrich-Alexander-Universität Erlangen-Nürnberg

On 12/29/2021 05:54 AM, Matthew Guberman-Pfeffer wrote:
> Dear Amber Community,
> I need to parameterize modified amino acids. As a reference, I tried to learn the procedure by applying the below steps to derive the charges for Tyr, but I get charges very different from what is given in FF14SB. Am I doing something wrong? If not, how can I make comparisons between simulations using the FF14SB charges for Tyr and simulations with my modified Tyr, if the procedure for the latter can’t reproduce the charges for the former?
> Step #1: Build ACE-TYR-NME capped peptide with LEaP.
> Step #2: Generate alpha-helix and beta-sheet Conformers by setting phi/psi angles with CPPTRAJ
> Step #3: Convert mol2 from CPPTRAJ for each conformer to Gaussian input and run with the route line: #p opt=modredundant freq hf/6-31g(d) pop=mk iop(6/33=2,6/42=6)
> where I’ve defined the phi/psi angles to be fixed at the canonical alpha-helix or beta-sheet values.
> Step #4: Extract ESP data from Gaussian output with espgen and concatenate results for alpha-helix and beta-sheet conformers.
> Step #5: Run residuegen with the below input file, in which I’ve specified the charges of the backbone N, H, C, and O atoms to be the values found for TYR in ff14SB.
> ATOM_CHARGE N1 -0.4157
> ATOM_CHARGE H4 0.2719
> ATOM_CHARGE C11 0.5973
> ATOM_CHARGE O3 -0.5679
> Below is the comparison of the charges in ff14SB versus what I obtain:
> TYR (ff14SB TYR (residuegen)
> -0.4157 -0.4157
> 0.2719 0.2719
> -0.0014 -0.0942
> 0.0876 0.1087
> -0.0152 0.7629
> 0.0295 -0.2468
> 0.0295 -0.2468
> -0.0011 -0.0982
> -0.1906 -0.0371
> 0.1699 -0.0366
> -0.2341 -0.0314
> 0.1656 0.0017
> 0.3226 0.1955
> -0.5579 -0.2451
> 0.3992 0.1849
> -0.2341 -0.0314
> 0.1656 0.0017
> -0.1906 -0.0371
> 0.1699 -0.0366
> 0.5973 0.5973
> -0.5679 -0.5679
> Please let me know if you need more information to help me with this matter.
> Best,
> Matthew
> _______________________________________________
> AMBER mailing list

AMBER mailing list
Received on Wed Dec 29 2021 - 02:00:02 PST
Custom Search