Re: AMBER: Thermodynamic Integration for conformational changes?

From: Gustavo Seabra <gustavo.seabra.gmail.com>
Date: Tue, 21 Aug 2007 23:33:36 -0400

On 8/21/07, Francesco Pietra <chiendarret.yahoo.com> wrote:
> I have carried out QM-MM (AM1) simulations for different conformations of a
> 98-atoms non-polymeric neutral molecule (solute) in explicit solvents.

Unfortunately, that doesn't tell us much. What kind of calculation did
you do? Did you use Amber? Was it a minimization or molecular
dynamics? Or was it just a single point calculation for each
structure?

> Would like to compare with DFTB, if I get a license for that.

About that, Marcus Elstner is out of the country (his country) right
now, so it's unlikely you'll get an answer from him soon. But you
could also try to download the parameters from www.dftb.org. You just
have to fill a registration form, then fax it to the number in the
form. Then you'll get an username and password, which you can use to
download the parameters. The specific set you want is 'mio-0-1', which
is the 'organic' set. The others contain atoms not supported in
Amber9.

When you get the parameter set from them, you will need to rename the
files. Their naming scheme is, for example, C-C.skf (Upper case,
'dash' and .skf extension). Then, all you have to do is to copy those
files to the $AMBERHOME/dat/slko directory, and rename them to
something like 'cc.spl' (lowercase, no dash, .spl extension), and that
should do.

> It is unclear to me how to relate above results to conformational energies, in
> particular how to deal with ESCF. The results seem to suggest that
> perturbations by the solvent cancel out by comparing ESCF for the conformers,
> though the theory behind has not been examined.


ESCF only gives you the energy of the QM region, which is influenced
by the presence of the MM atoms. The energy of the MM atoms, however,
is *not* included there. If you want to look at the energy of the
*system*, what you really want is the total energy instead, which does
include the solvent molecules.

> At any event, I would like to calculate free energy changes for this system.
> The question whether thermodynamic integration is appropriate for complicated
> conformational changes received authoritative NO at Amber8's time (archives)
> and is probably to be extended to Amber9. Are umbrella sampling or MM_PBSA
> potential solutions? The first one, from the manual, also seems unable to deal
> with multiple dihedral changes from one conformation to the other one. What
> about MM_PBSA? Is any example to which to refer for the evaluation of absolute
> free energies in explicit solvent by MM_PBSA?

I'll not pretend to understand much here, and I'd advise you to take
whatever I say with a whole pound of salt, but notice that you have a
really big system and, for what I can imagine, some big conformational
changes from one point to the other. Have you tried to imagine what
would you use as a reaction coordinate? The big problem for you (I
believe) is that for most free energy methods you need to define not
only initial and final states, but you also need some idea about the
reaction path you want to probe. AFAIK, there's no 'magical' method
that will get initial and final point and locate the minimum free
energy path between them. And that would basically be the reason those
methods won't work for you: with that many dihedrals, how many
different reaction paths can you think of? When can you be satisfied
that you have finally found the lowest energy one? (But I can be wrong
here... anybody?)

In principle, you can also try for example a real long REMD
simulation, until you see enough transitions beteween the two states
that you can consider the populations to be equilibrated. (Follow the
populations of the conformers until they stabilize) Then, get the free
energy difference from the population of each conformer. But that
would probably take longer than what you (or anyone I know) would be
willing to wait if you are going to do that with QM.

But I don't mean to be negative here. Hopefully, someone with more
experience will chime in here with a better solution.

Good luck,
Gustavo.
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Wed Aug 22 2007 - 06:07:49 PDT
Custom Search