From: Jérôme Hénin (jhenin_at_ifr88.cnrs-mrs.fr)
Date: Fri May 04 2012 - 09:32:29 CDT
Hi Xiaohu,
Some of the issues you are having could be due to a bad reaction
coordinate. You restrain the center of mass of the water slab, but
given its size, the surface of the water is still probably able to
move around a little with moderate distortion of the slab. So for a
given value of the reaction coordinate, the surface can move back and
forth and adapt to the ion's position. This probably makes the process
harder to sample, because you have to sample large fluctuations of the
whole slab instead of small local fluctuations of the interface. That
could probably be diagnosed with a detailed inspection of the
trajectory in VMD.
Of course, there is no simple coordinate to describe the actual
distance of the ion to the water interface. Still, if you follow JC's
advice and use a smaller slab, the precision of your RC will improve.
Best,
Jerome
On 4 May 2012 15:51, JC Gumbart <gumbart_at_ks.uiuc.edu> wrote:
> Yes. But I still question the need for such a large slab to look at just
> one ion.
>
>
>
> From: Xiaohu Li [mailto:xiaohu-li_at_northwestern.edu]
> Sent: Friday, May 04, 2012 8:49 AM
> To: JC Gumbart
> Cc: namd-l_at_ks.uiuc.edu List
> Subject: Re: namd-l: forces saved in forcedcd
>
>
>
> Thanks, JC.
>
> I recall that ABF also uses the colvar subroutine, which means that it is
> also un-parallelized?
>
>
>
> Xiaohu
>
>
>
> On May 4, 2012, at 12:58 AM, JC Gumbart wrote:
>
>
>
> My gut feeling is that the separate restraints are contaminating your PMF.
> Why do you need such a large slab?
>
>
>
> The ABF tutorial gives an example of what you're trying to
> do: http://www.ks.uiuc.edu/Training/Tutorials/namd/ABF/tutorial-abf.pdf
>
>
>
> See if you can reproduce those results and then modify things one at a time
> from there to determine where the potential error is appearing.
>
>
>
> On May 3, 2012, at 10:40 PM, Xiaohu Li wrote:
>
>
>
> I first tried to use the colvar option to apply a harmonic constraint
> between the center of mass of the liquid slab and the ion, however, since my
> slab has tens of thousands of atoms, the un-parallelized colvar code slow
> down my
>
> simulation by a factor of ten!
>
> So instead, I applied a very large harmonic constraint on the COM of the
> slab using the SMD option and another harmonic constraint on the ion using
> regular harmonic constraint.
>
> Of course, the slab has been equilibrated.
>
> Another thing is that when I inserted my ion into the slab, I used VMD to
> deleted A solvent molecule nearby to avoid any close contact. I'm guessing
> this effect is small since I have like 700 solvent molecules.
>
> I ran each window dynamics for 600ps, where the first 100ps was discarded as
> equilibration.
>
>
>
> Xiaohu
>
>
>
> On May 3, 2012, at 10:05 PM, JC Gumbart wrote:
>
>
>
> It's easier to set up, but should be equivalent in the end.
>
>
>
> How did you prepare and run your US sims? I cannot for the life of me
> figure out why you are getting an oscillatory pattern from US.
>
>
>
> On May 3, 2012, at 9:57 PM, Xiaohu Li wrote:
>
>
>
> I have used umbrella sampling by either using WHAM or averaging the forces
> the the window.
>
> How's ABF compared to those approaches? Are there any qualitative
> differences?
>
>
>
> Xiaohu
>
>
>
> On May 3, 2012, at 8:11 PM, JC Gumbart wrote:
>
>
>
> How are you calculating the PMF? Did you try, say, ABF?
>
>
>
> On May 3, 2012, at 7:51 PM, Xiaohu Li wrote:
>
>
>
> Hi, Chris,
>
> Thanks for taking time for pondering on my question.
>
> In my case, I would have to stick with the NVT simulation since I'm
> using the drude polarizable force field. The reason I ask is because I'm
> doing a PMF of ions pulled out of bulk(glycerol), which the reaction
> coordinates is quite simple as the surface normal direction. I'm seeing
> something quite difficult to explain since from a very naive understanding,
> the PMF should be flat deep in the bulk(as seen in water, methanol, etc...).
> I'm getting quite oscillatory pattern in the bulk and the i'm talking about
> several kcal/mol oscillation(using either wham or the direct forces). That
> leads me to this question. Something I notice is that the frequency of the
> Langevin friction affect the self-diffusion. So extrapolate from that, it
> should affect the forces.
>
> Any more opinions on that?
>
>
>
>
>
> Xiaohu
>
>
>
> On May 3, 2012, at 6:42 PM, Chris Harrison wrote:
>
>
>
> Yes. However, in the limit of converged sampling the Langevin forces are
> theoretically canceling out to noise. ...theoretically. In practice, it is
> entirely possible that when using a Langevin thermostat you might see the
> periodicity produced from the kicks when analyzing the forces.
>
>
> "Will the Langevin forces affect PMF calculations?" That entirely depends
> on the PMF you are performing. Many considerations go into answering that
> question. Consider: Are the forces being predominately measured in the PMF
> of a small enough scale that the Langevin kicks can be resolved? If
> performed correctly, the Langevin forces should look like noise. Another
> consideration: a purist would argue to perform an NVE simulation, removing
> as many external forces/perturbations from the system as possible.
>
>
>
> Unfortunately the answer to your question is: yes, they can affect PMFs, but
> a) will it matter for the question you are asking, and b) you will have to
> test to actually determine for sure; or alternatively pursue an NVE ensemble
> simulation.
>
>
>
> Best,
>
> Chris
>
>
>
>
>
>
>
> On Thu, May 3, 2012 at 3:43 PM, Xiaohu Li <xiaohu-li_at_northwestern.edu>
> wrote:
>
> Dear NAMD users,
> I have a question regarding the forces saved in forcedcd file.
> suppose I'm doing NVT using Langevin dynamics, will the forces saved in
> forcedcd contain the random forces from Langevin dynamics?
> Do those random forces affect, say, PMF calculations if they are
> included?
> Thanks.
>
>
>
>
>
> --
>
> Chris Harrison, Ph.D.
>
> NIH Center for Macromolecular Modeling and Bioinformatics
>
> Theoretical and Computational Biophysics Group
> Beckman Institute for Advanced Science and Technology
> University of Illinois, 405 N. Mathews Ave., Urbana, IL 61801
>
> http://www.ks.uiuc.edu/Research/namd Voice: 773-570-6078
>
> http://www.ks.uiuc.edu/~char Fax: 217-244-6078
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
This archive was generated by hypermail 2.1.6 : Mon Dec 31 2012 - 23:21:30 CST