Re: [AMBER] DIELC constant for normal mode analyses in vacuum ?

From: Marek Maly <>
Date: Fri, 24 Jun 2011 17:56:37 +0200

Thanks a lot again !

My motivation/goal is here to obtain some rough estimates of free
solvation energies
(at least for purpose of relative/qualitative comparison of similar
molecules (branched polymers)) based
on separate MD simulations (in vacuum, in water) for each molecule and
consequent G analyses in both states
(water, vacuum).

Approximative formula which I assume is: dG_solvation =
G_solute_in_solvent(including interaction energy with solvent ) -

(solute-water interaction calculated using implicit solvent)

I know that I could adopt more simple approximation ( dG_solvation =
solute interaction energy with solvent ) but this is
OK only for "rigid" molecules which do not change their conformations
and/or state (protonation due to solvation) due to solvation. In my case
this change is really
NON-negligible so this easiest way could not be used in my opinion so I
need to include G_solute_in_vacuum.

So the question is if the errors in vacuum calculations (wrong molecular
configurations obtained from MD in vacuum , enthalpy - here just solute
bond/nonbond energy , entropy )
might be so high that even for my comparisonal purposes my above mentioned
approach is not possible using Amber.

I also know that "point of using either GB or epsilon=4r is *not* to
simulate the system in vacuum, but to have an implicit solvent model" I
was just asking
if the parameters of these models might be changed to reflect better
calculation in vacuum (e.g. to change EXTDIEL parm. in .NM section from 80
to some smaller value
in case of using GB model with NAB) or to decrease DIELC parm again in .NM
section in case of using distance-dependent dielectric constant model ...

I know that the best answers I could obtain are from my own "experiments"
using experimentally already analyzed molecules, but I just have no time
for this at this moment ...

I was also asking for some more details regarding your general sentence
"But you should remember that the potentials were
parameterized for condensed phase simulations, and often do a very poor
job of representing what would really happen in vacuum."

So could you please more comment this ? At least about which potentials
you are speaking ( forcefield ? or el. potentials in GB ? )

Thank you in advance very much for your final explanation and time !

     Best wishes,


Dne Fri, 24 Jun 2011 04:35:40 +0200 David A. Case
<> napsal/-a:

> On Thu, Jun 23, 2011, Marek Maly wrote:
>> So could you say some recommendation about the DIELC parm. from .NM
>> section regarding normal mode analyses in vacuum ?
> For the purposes of getting a rough estimate of harmonic entropies,
> using the
> default values is probably's been used a lot. But I personally
> would
> use the python version, which uses GB theory to get normal modes. These
> seem
> to provide strutures that are (to a small but non-negligable extent)
> closer to
> solution structures than those that use a 4r dielctric constant.
> Note that the point of using either GB or epsilon=4r is *not* to
> simulate the
> system in vacuum, but to have an implicit solvent model, that is, one
> that has
> no explicit solvent molecules, but which nevertheless mimics the
> behavior of
> the solute in solution. So, it just "looks like" a vacuum simulation,
> but the
> goal is to get an estimate of solvated properties. [So I really
> misunderstood your goal in my previous reply.]
> ...hope this helps...dac
> _______________________________________________
> AMBER mailing list

Tato zpráva byla vytvořena převratným poštovním klientem Opery:
AMBER mailing list
Received on Fri Jun 24 2011 - 09:30:04 PDT
Custom Search