Thank you, that was a very clear answer.
I actually used igb=1 (because almost all the tutorials on the Amber
website use this setting). I will change it for igb=8 in the future.
"
GB simulations are infinitely dilute. It doesn't matter how strongly two
molecules bind, basic stat thermo tells you that the bound state of an
infinitely dilute solution is not stable, regardless of binding strength.
Remember that free energies of binding have an entropic component, and as a
result depend on concentrations (which are really just probabilities).
If the system*starts* bound, then a strongly bound complex will take a
long time to dissociate, but expecting the small peptide to "capture" the
ligand molecule in implicit solvent may be overly optimistic, unless it
starts out "almost" captured.
By contrast, explicit solvent calculations have a well-defined (and often
high) concentration, so you can actually expect to see binding events
sometimes (D. E. Shaw's research group has done some of this)."
Ok, I didn't think at all about the dilution. So basically, when you are
trying to see binding events, the general recommendation is to use
explicit solvent.
Do I understand well ? So binding events -> definitely no implicit solvent ?
In my case, I can't start with an "almost" captured ligand, because I'm
trying to find out how the two molecules bind.
Le 04/11/2015 15:01, Jason Swails a écrit :
> On Wed, Nov 4, 2015 at 8:42 AM, Jean-Patrick Francoia <
> jeanpatrick.francoia.gmail.com> wrote:
>
>> Hello,
>>
>> I'm trying to do MD on a system made of a short peptide and the molecule
>> of cocaine. It is just a training. The peptide is a hexapeptide
>> (QHWWDW), easily generated with the 'sequence' command of xleap. The
>> library for the cocaine residue was built using antechamber.
>>
>> I merged the peptide and the cocaine. I performed a short minimization
>> using implicit solvent, and then, I tried to run a molecular dynamic. It
>> "worked", meaning the MD completed without errors. I now have a splendid
>> trajectory (1 ns). But I don't know what it means, and what to expect.
>>
> "Using implicit solvent" is about as specific as saying "using the Amber
> force field". There are many implicit solvent models, some more accurate
> than others (by a wide margin). The most modern GB model, and the one that
> is probably the most accurate compared to explicit solvent calculations, is
> igb=8 with mbondi3 radii. I would recommend using this combination.
>
>
>
>> I know the peptide and the cocaine bind "strongly" (I can't quantify
>> strongly). There are experimental evidences. So I was expecting the
>> peptide would fold and catch the cocaine, and that at some point, the
>> complex would stop changing. I was hoping to see a particular
>> conformation of the complex. Instead, I see the two molecules
>> interacting, but no preferred conformation (ok, maybe one, appearing
>> sometimes), but nothing really fixed.
>>
>> Was I right to expect that, or not ?
>>
> GB simulations are infinitely dilute. It doesn't matter how strongly two
> molecules bind, basic stat thermo tells you that the bound state of an
> infinitely dilute solution is not stable, regardless of binding strength.
> Remember that free energies of binding have an entropic component, and as a
> result depend on concentrations (which are really just probabilities).
>
> If the system *starts* bound, then a strongly bound complex will take a
> long time to dissociate, but expecting the small peptide to "capture" the
> ligand molecule in implicit solvent may be overly optimistic, unless it
> starts out "almost" captured.
>
> By contrast, explicit solvent calculations have a well-defined (and often
> high) concentration, so you can actually expect to see binding events
> sometimes (D. E. Shaw's research group has done some of this).
>
> HTH,
> Jason
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Nov 05 2015 - 02:00:03 PST
