From chemistry-request %-% at %-% ccl.net Thu Jun 20 03:51:55 1991
Date: Thu, 20 Jun 1991 09:34:56 +0200
From: Finn Drablos <finn.drablos ^at^ sintef.no>
Subject: Re: Free-energy perturbation calcs of zinc<->peptide binding
To: chemistry*- at -*ccl.net
Status: R

==================

>Hello to all AMBER gurus out there:
>
>I want to do a slow-growth free-energy perturbation calculation to estimate
>the Gibbs free-energy change for the process zinc-ion + peptide -> peptide.
>
>I have parameterised the zinc-ion in AMBER 3.0A as an 'H-bonding' type 
>of atom ie. it has 10-12 well-depth and equilibrium bond distances with
>all relevent donor atoms on the peptide (peptide C=O, carboxylate-O,
>amine H2N's etc).  In addition, it has a point-charge calculated by fitting
>the electrostatic potential with MOPAC.  Thus the Zn<->donor atom interaction
>energy has two major sources (i) electrostatics and (ii) 10-12 well-depth.
>
(...)
>
>Any comments about the treatment of the zinc-ion <-> peptide binding ?
>
>Looking forward (I think) to some discussion ....
>
>----
>Alan Arnold             	     |  Phone: +61 62 68 8080
>Chem. Department,University College  | ACSNET: apa \\at// ccadfa.oz 
>Australian Defence Force Academy     |   UUCP: ...!seismo!munnari!ccadfa.oz!lpb 
>CANBERRA  ACT 2600 Australia         |   ARPA: apa%ccadfa.oz(+ at +)SEISMO.CSS.GOV

I am not an AMBER guru, but I have used Discover for some peptide/Zn
simulations, and have a few comments.

In general I think that most molecular mechanics programs do have
problems with metal ions, and two important reasons for this 
are polarization effects and
coordination. With a metal ion we introduce a strong point charge into
the system. Well, not all ions should be handled as point charges,
really. But at the current level of MM theory they are, in most cases.
Anyway, this will introduce polarization effects in groups close to
the ion. This is important, especially in dynamics simulations, as
polarization is a dynamic property.

But most MM programs don't include polarizability. I have heard that
AMBER 4.0 includes polarization effects, but I don't know how they
are represented in the forcefield (Any comments, out there?). But
maybe you should try to get 4.0 before you start to do simulations.

The other problem is coordination. Some ions like Ca++ don't care
really, some like Ni++ are very sensitive, and Zn++ is somewhere
in between. But the problem is that you really need to modify you
forcefield in order to include coordination effects, see for
example A. Vedani, J. Comp. Chem., 1986, 7, 701-710.

And one last potential problem. I don't have the paper on AM1 
parameters for Zn at hand, but at least the MNDO parameters where
mainly estimated on the basis of organometallic compounds, like
ZnEt2 and ZnMeI, and not complexed ions. So you should verify that
the result is reasonable (compare the geometry to known X-ray
structures) (More comments on this, out there?).

But as you may have guessed, I ended up doing my computations
(only minimization / simple dynamics) very much the same way as
you have indicated. I didn't have much choice, did I ... But be
careful when you interpret you results!

Regards,

Finn Drablos               PHONE +47 7 997710  FAX +47 7 997708
MR Center, SINTEF UNIMED,  drablos;at;marvin.mr.sintef.no
N-7034 TRONDHEIM, NORWAY   finn.drablos |-at-| sintef.no 
---------------------------------------------------------------