Hello,
Gustavo is right. Different initial conditions give far different results
in non-converged MD simulations (i.e. simulations that were not run for long
enough). You could probably even start 3 separate simulations from the same
starting point, change the random seed for langevin dynamics, and obtain
these "drastically" different results for RMSF/RMSD/energy/etc. amongst
those three seemingly "identical" simulations.
I haven't actually ever tried that exercise, but such is the nature of
chaotic systems.
Good luck!
Jason
On Wed, May 26, 2010 at 10:11 AM, Gustavo Seabra
<gustavo.seabra.gmail.com>wrote:
> Hi Reza,
>
> On Wed, May 26, 2010 at 9:41 AM, M. Reza Ganjalikhany wrote:
> > I have problem with adjusting the number of minimization steps for my MD
> simulation.
> > I ran 3 MD simulations in explicit solvent with a protein containing 181
> a.a. for 2 nanosec and I got
> > 3 different sets of data in which the RMSd, RMSf, radius of gyration and
> > other data were a bit different in each one!!
> >
> > Minimization steps were adjusted as below:
> > run #1 : 2500 steps (1000steps steepest decent+1500 steps conjugate
> gradient)
> > run #2: 5000 steps (2000 steps steepest decent+3000 steps conjugate
> gradient)
> > run #3 :10000 steps (2000 steps steepest decent+8000 steps conjugate
> gradient)
>
> You mention running 3 MD simulations, but the information you put here
> is for minimizations. Can you clarify what you actually did?
>
> > Is it normal to obtain different sets of data upon doing different
> minimization steps?
>
> I'd say yes, it is normal. The potential energy surface for systems
> this big is very complex, with many, many minima. A small change in
> the minimizer parameters could easily take you to a different minimum,
> and get trapped there.
>
> > How many minimization steps should be sufficient? How is it possible to
> find
> > ot its suficiency?
>
> If you are trapped in different local minima, no number of steps will
> help. Even if you could get all 3 results to the same structure, there
> would be no guarantee that you are in the global minimum.
>
> Is finding a global minimum really that important for your project? If
> so, the best you can do is probably changing the minimizer to LMOD (se
> manual). However, as Dr. Steinbrecher mentioned, the super-tight
> minimization becomes just a waste of time if you intend to run MD with
> it later. If the results you show in the pictures are from MDs run
> **after** the minimizations you mention, then all it tells me is that
> the runs are not converged yet. When converged, you should get the
> same thermodynamic properties from any of the simulations.
>
> Cheers,
>
> Gustavo Seabra
> Professor Adjunto
> Departamento de Química Fundamental
> Universidade Federal de Pernambuco
> Fone: +55-81-2126-7417
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
--
Jason M. Swails
Quantum Theory Project,
University of Florida
Ph.D. Graduate Student
352-392-4032
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed May 26 2010 - 07:30:37 PDT
