Re: [AMBER] Clustering VS constant rmsd

From: vaibhav dixit <vaibhavadixit.gmail.com>
Date: Wed, 17 Oct 2012 23:17:02 +0530

Dear Catein,
I have also recently started to use the clustering algorithms implemented
in Amber. So I would myself like to hear from the experts. Nevertheless
following is what I understand about the questions that you have raised.

On Wed, Oct 17, 2012 at 10:53 PM, Catein Catherine
<askamber23.hotmail.com>wrote:

>
> Dear AMBER users,
> When we finish the MD simulation, we got a large number of mdcrd files.
> In the old days, we check the rmsd change over time, when the rmsd become
> constant, we define the system is converged, and production period start
> where we can do the ptraj (hbond, average pdb, etc....).
> Recent trends seems to be doing clustering analysis instead. We don't
> need to check the rmsd changes over time. Sometime, we just do multiple
> simulations with different starting structures.
>
Checking rmsd is an essential part of the MD trajectory analysis and it
helps us to validate if the system has reached equilibrium (converged) i.e.
expected to be closer to what would a real protein/DNA/any other system
would be. As given in the analysis portion of the amber first tutorial we
should also verify whether "energy" "pressure" and "volume" of the system
have stabilized or not (this may depend on the NVT, NPT ensemble employed
for the MD). http://ambermd.org/tutorials/basic/tutorial1/section5.htm

Based on your experience, when we do the clustering analysis, do we still
> need to check if the system is converged? If yes, what is the right
> approach?
>
We should first check for rmsd and other parameters before doing
clustering. If the system has a very dynamic nature (e.g. nuclear
receptors/polypeptides) small rmsd values may not be a necessary criteria
and a thorough understanding of the dynamical behavior of the system is
necessary before making such decisions. As mentioned above energy, pressure
and volume of the system must have converged before performing analysis.

> If we are doing the long simulation, should we still do the clustering
> analysis only for the production period? How about the equilibrium period?
> Should we also include those mdcrd files starting from time zero?
>
I think the equilibrium portion should be omitted because it is that part
of MD where the program is adjusting the system so that it can represent
the real world system as close as possible. I hope this helps.


> Best regards,
> Catherine
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>



-- 
With regards
Vaibhav A. Dixit
Ph.D. Scholar
Department of Medicinal Chemistry
Natl. Inst. Pharm. Edu. & Res. (NIPER)
Sector 67, Phase X,  S.A.S. Nagar (Mohali)
Punjab -160 062 INDIA
Phone (Mobile): +919915214408
E-mail: vaibhavadixit.gmail.com
www.niper.nic.in
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Oct 17 2012 - 11:00:04 PDT
Custom Search