From: Ziemys, Arturas (Arturas.Ziemys_at_uth.tmc.edu)
Date: Tue Jun 23 2009 - 13:04:24 CDT
Yes, I have noticed it also: the accumulation of statistics at the interface. I solved it to some extend splitting RC close to the interface by objecting the molecule go close and be 'adsorbed'.
I suspect several things those could be responsible together or alone. One is that is there is ordered solvent effect at the interface that probably can 'cage' your molecule at the interface. Another my thought is that the adsorbed state is in energetically very steep energy surface that leads to high energy fluctuations, therefore it takes just a lot to reach the true average of force. However, it at some point force converges at those bins close to the surface, but molecule is not escaping. If someone will be able to reveal the true nature of this that would be very great. Increasing constant should not help, because it acts just on the ends of RC.
Division of NanoMedicine
Department of Biomedical Engineering
The University of Texas Health Science Center at Houston
1825 Pressler Street, Suite 537
Houston, TX 77030
Phone: (713) 500-2311
Email : Arturas.Ziemys_at_uth.tmc.edu
From: owner-namd-l_at_ks.uiuc.edu [owner-namd-l_at_ks.uiuc.edu] On Behalf Of Sumanth Jamadagni [jamads_at_rpi.edu]
Sent: Tuesday, June 23, 2009 11:32 AM
To: Sébastien Légaré
Subject: Re: namd-l: ABF simulations
Dear Sebastian and NAMD users,
Thanks for your suggestion. I'll try using a larger force constant. But
I was under the impression that the force constants are used primarily
to ensure that the molecule (or group of atoms) stays within Ximax and
Ximin, (within which it remains, but doesn't sample the rest of the
desired reaction coordinate region, just one end) . We have also tried
splitting the region of interest into smaller regions, but the same
problem exists even for much smaller regions (~3 Angstroms).
I am not sure if anybody else if facing the same problem when using zCoord.
Else, we have to go back to umbrella sampling which is what we were
trying to avoid.
Sébastien Légaré wrote:
> Dear Sumanth and NAMD users
> I do not have a final answer to this question but we have had the same
> problem. In most cases, we used a distance reaction coordinate between two
> amino acids in a protein. Since the amino acids are constrained by the rest
> of the protein and do not sample a larger volume at larger distances, I
> suspected that the problem came from a too large Jacobian correction term. If
> you used the zCoord reaction coordinate there is no Jacobian term and I still
> do not see why this happens. You can try setting forceConst to 100
> kcal/mol/Angstroms^2 as suggested in page 3 of Chipot, Hénin, J. Chem. Phys.
> 123, 244906 (2005).
> Sébastien Légaré
> On June 9, 2009 10:08:10 pm Sumanth Jamadagni wrote:
>> Dear NAMD users,
>> We have been trying to run some simple ABF simulations to calculate to
>> PMF along z between a molecule and a surface (or similar systems, we
>> have faced the same problem).
>> The problem is this: we specify an /ximax/ and /ximin/ for the reaction
>> coordinate. The center of mass of the molecule initially samples the
>> reaction cordinate and then the molecule is stuck at one end of the
>> reaction coordinate (ximax or ximin) and doesn't diffuse and sample the
>> whole range any more. It therefore accumulates statistics only in the
>> first or last few bins, and there are very few statistics in the rest of
>> the bins. I will try running the sims for much longer to see if that
>> helps, but I suspect I am missing something.
>> Sumanth Jamadagni
> Sébastien Légaré
> Ph. D. Student
> Laval University
> Quebec, Canada, G1V 0A6
> 418-656-2131 #11577
This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:51:06 CST