RE: AMBER: Gas-phase energies (and more)

From: Yong Duan <>
Date: Sun, 28 Jan 2007 19:43:56 -0800

Dear Jason,

At the 20-50 aa level, you are dealing with a potentially very demanding
problem. But if you are interested in qualitative trend, given the potential
cost of the full polarizable model with explicit solvent, you may explore
the fixed charge models with GB. Again, a potential problem could arise
particularly in the gas phase and the electrostatic interactions at the
surface could be somewhat exaggerated. Clearly, I need to do my homework to
comment more. But, intuitively, I'd say the problem might not be as bad as
it sounds because although the charges were developed in "condensed phase"
and the surface of the peptides is fully exposed to vaccum, the surface
atoms are still partially surrounded by other atoms (so in a "solvent"). As
a qualitative study of the conformational stability, this could be a
reasonable approach.

I also agree with Carlos' comment that ff99 is probably not a good choice.

As for other questions regarding conversion between z-mat and other formats,
I am not an expert in that area. But, as Dave said, if you post your input
files someone might be able to look into it for you more carefully.


-----Original Message-----
From: [] On Behalf Of
Jason K
Sent: Friday, January 26, 2007 1:36 PM
Subject: Re: AMBER: Gas-phase energies (and more)

Dear Yong (and everyone else who might be reading this),

Thank you very much for your fast reply.

The truth is that I am only using the R(+)-A4 pentapeptide as a test system
because sufficient data using experimental techniques very similar to ours
have been published. What I am actually interested to investigate (in very
few words) is the stability of particular conformations of peptides in
different environments (vacuum, organic solvents and water, the latter being
either implicitly or explicitly simulated). The systems that I simulate
range from 20 to 50 amino acids, and the total simulation time amounts to
several hundreds of nanoseconds. Therefore computational time is an issue. I
must confess however that I have not done any benchmarks on my systems using
polarisable force fields; I suppose this is the next thing to do.
So far I've been running everything using ff99, but I have been thinking for
several months it may be worth switching to "better" parameter sets.

Now, to a few more basic questions...

Are there any "built-in" tools and/or other scripts one would like to share
for performing a potential energy surface scan, so that I could compare
simple PES maps from MM from QM calculations?

If not, could someone point me to the direction of performing MD or energy
minimisation using internal coordinate restraints? Are the options used the
same used for NMR-type restraints?

Finally, does any software exist that can reliably convert from amber
restart (given the parameter file) to internal coordinates (z-matrix) AND
back? I have tried with antechamber, but is generally seems a bit slow (not
surprisingly since it's prime role is parameter fitting and not mere format
conversion) and often crashes out. Furthermore, it does not seem to convert
directly to restart files but only to pdb (a lot of time being spent to
creating unnecessary lists of atom and residue numbers, residue and chain
names) and not always reliable: it worked with the very simple Ace-Ala-Nme
structure, but with my 20-aa-long peptide:

*** glibc detected *** double free or corruption: 0x00000000005a8490 ***
Segmentation fault

and one other time, on a perhaps slightly more exotic heparin molecule it
projected all coordinates to a single axis (!). As far as I know OpenBabel
does not cope with the conversion either, so does anyone know any software
that can do this? I know it is my general ignorance on scripting that is
probably to blame here, yet if there is indeed a script that can be relied
on this conversion, for a system of any size, I would feel very obliged.

[I know that for simple PES scans on small molecules I could extract the
internal coordinates from the gaussian output, convert them into pdb with
babel or newzmat or something similar and finally into rst format, however
this coordinate conversion is a more general problem and the rst to pdb to
zmat to pdb to rst procedure seems quite cumbersome, slow and error-prone].

I do not want this to sound like a criticism, I really appreciate
everybody's help, just some simple issues can take more time than they ought
to. Unfortunately, against a computer I still feel defenseless.

Best Regards,


On 1/25/07, Yong Duan <> wrote:


Actually the term "gas-phase" or "condensed phase" is somewhat relative.

Taking your system as an example, if the charges are developed using the QM
esp of the entire molecule, it is probably justified to use gas-phase esp if

the intention is to model its gas-phase behavior, assuming you use multiple
representative conformations in the esp calculation. However, if the
approach is to obtain the charges of the di-peptides and piece together the
charges for the whole molecule, it is not a clear cut as to how to obtain
the charges. The reason for this is that, as you might anticipate, the
peptide (Ace-Arg(+)-Ala-Ala-Ala-Ala-Nme) is sufficiently large that it may
have tendency to form compact (collapsed) structures, even in gas-phase. If
this is the case, the residues that tend to be burried should perhaps use
the condensed-phase charges. Intuitively, only those exposed atoms should
assume gas-phase charges and other atoms should behave like in some sort of
solvent. This is not exactly "hair-splitting". I would say that only the
fully polarizable force field can do this (or, in a more modest tone, has
the potential to do the job).

If you only want to simulate these small peptides in gas-phase, I think you
should use the ff02 charges with polarizability because, theoretically, it
should be able to deal with it. In fact, if you go through the paper of
Cieplak et al (JCC paper with Kollman and Caldwell), the charges of ff02 was
fitted in gas-phase with polarization. If you are concerned by the added
cost, simulating such small peptides is really pretty easy these days even
using the fully polarizable force field.

As for the torsion parameters, because each torsion parameter set is closely
related to the charges, it makes sense to try Wang et al's torsion
parameters first. In the work of Wang, the torsion tuning was done in
gas-phase. In other words, the fitting of the torsion was done by comparing
the di-ala gas-phase energies. But the tests were done in aqueous solution.
So, it should work.

Hope this helps.

Good luck!


-----Original Message-----
From: [] On Behalf Of
Jason K
Sent: Thursday, January 25, 2007 3:04 PM
Subject: AMBER: Gas-phase energies (and more)

Dear AMBER users,

Much of the research in the group where I work addresses the structures of
boimolecules in a solvent-free environment. We are trying to reproduce our
experimental results by molecular mechanical simulations and -subsequently-
the concern that most AMBER parameter sets are tuned to replicate
condensed-phase energies for small peptides becomes quite relevant. Since
the electrostatics contribute the most to the energy of the system
(especially in vacuo), also being the kind of interactions that differ the
most between the gas phase and solution, I would like to know what approach
I could adopt to generate charge assignments that are more likely to mirror
the electrostatics in the absence of solvent. So far (to partially answer my

own question) I have come across two approaches, (excluding, of course
re-parameterising the force field in the gas-phase):

1) Use the "ff02" residues but with ipol=0 (thus using the "gas phase",
"static" point-charges). [From what I have gathered, a polarisable force
field should handle the electrostatics in a vacuum, implicit or explicit
solvent, at least in theory, but I do not know to what extent this is true
for the models in AMBER or to what extent they have been tested and hence I
am unwilling to make this assumption]. I am still unsure on whether this is
indeed appropriate, what objections could one raise against using ff02
charges for gas-phase calculations?

2) "Scaling down" the HF/6-31G* charges (by e.g. a factor of 0.8) to
somewhat relieve the overestimated dipoles. I would be extremely grateful if
someone could tell me how to calculate the partial charges for residues with

non-integer charge. Has this approach been tested?

Does anyone know any other methods for a "better" representation of
gas-phase electrostatics of biomolecules?

Finally, I am bound to ask about other parameters, especially dihedrals.
Will the tortional terms in parm99.dat or even frcmod.02 (dot
somethingsomething) drive the peptide away from the "real" gas-phase minima?
I am currently running calculations on a small system
(Ace-Arg(+)-Ala-Ala-Ala-Ala-Nme) which has been addressed previously and see

which parameter set of the ones available gives "better" energies compared
with QM, or correspondence with experiment. Yet if anyone could propose me a
different testing method or a more "standard" system, please do so. (The
reason a protonated peptide is needed is that neutral peptides are not
directly observable by mass spectrometry and related techniques).

Thanks in advance


PS1. I haven't read the paper carefully yet, but in Wang et al. J. Comp.
Chem. (2006) 27(6):781-790 were both the MM and QM calculations performed in
water for the Ace-Ala-Nme dihedral scan ("figure 3") or was the QM done in
the absence of (implicit) solvation?

The AMBER Mail Reflector
To post, send mail to
To unsubscribe, send "unsubscribe amber" to

The AMBER Mail Reflector
To post, send mail to
To unsubscribe, send "unsubscribe amber" to
Received on Wed Jan 31 2007 - 06:07:41 PST
Custom Search