[gmx-users] Re: Resilin Project

[Justin A. Lemkul]

You've basically got the idea.  As a subtle point, the "strength" of a method is 
completely user-defined (by the tau_p parameter, how frequently the temperature 
is coupled).  No one method is "stronger" than another, but it is correct to say 
that the Nose-Hoover thermostat produces a rigorous NVT ensemble.  When used 
with Parrinello-Rahman for pressure control, a rigorous NPT ensemble is produced.

If I remember our discussion correctly, you want to fix one end of the molecule 
and pull on the other, correct?  If so, your pull_group0 will be the residue at 
one end of the protein (N- or C-terminus, depending on how you set it up), and 
pull_group1 will be the other terminus.

-Justin

Ravi Kappiyoor wrote:
> Thank you, Justin. To make sure I understood correctly, you use weak 
> coupling methods during equilibration (such as V-rescale and Berendsen) 
> in order to try and avoid large fluctuations. However, during actual MD 
> simulations, you would recommend using stronger coupling methods such as 
> Nose-Hoover, correct? Also, one question I forgot to ask in the last 
> e-mail, in the umbrella sampling, I believe you used the pull command to 
> pull chain A of the amyloid fibril with respect to chain B. In my case, 
> I will be testing a single chain of resilin, so how would I adapt this 
> segment?
> 
> On Wed, Aug 4, 2010 at 6:45 PM, Justin A. Lemkul <jalemkul at vt.edu 
> <mailto:jalemkul at vt.edu>> wrote:
> 
> 
>     Hi Ravi,
> 
>     Your questions raise a number of interesting points, most of which
>     take a few years of experience to fully appreciate :)  I will try to
>     give you a short take on the whole idea of equilibration, as best as
>     email will allow.  I will note, though, that to fully address the
>     subtleties of what you're asking, it takes quite a lot of textbook
>     reading.  I can recommend a few to you, if you like.  We devoted
>     several weeks to this type of discussion in the class we taught.
> 
>     The initial equilibration of any system should be approached gently.
>      Solvent that is added has a very regular, almost crystalline,
>     configuration.  It is very unnatural.  NVT equilibration is
>     typically done before NPT, although I happen to know that the
>     amyloid fibrils are incredibly resilient and do not require NVT
>     equilibration.  This is only something I happen to have figured out
>     after a few years of working on my project.  Generally speaking
>     (about 99% of the time), you should do NVT first.  Weak temperature
>     coupling methods like Berendsen or V-rescale are used because the
>     system is far from equilibrium and any initial clashes will induce
>     huge temperature fluctuations if you apply the Nose-Hoover method.
>      The only explanation for these observations lies in the primary
>     literature for the thermostat algorithms.  I'll warn you, the math
>     is heavy. Sorry to say there is no quick link to provide.  I've
>     never seen anyone be able to give a concise description of these
>     algorithms, let alone the difference between them :)
> 
>     I used Parrinello-Rahman for the lysozyme NPT since, during NPT, the
>     system is less sensitive to pressure oscillations once NVT has been
>     done.  In the umbrella sampling tutorial, since I did not do NVT, I
>     maintained a weak coupling method.  This is because the first
>     equilibration you do should always use some form of weak coupling.
> 
>     Certainly during actual data collection you should use Nose-Hoover
>     or V-rescale for your thermostat and Parrinello-Rahman for your
>     barostat.  The Berendsen method is quick and convenient, but does
>     not generate a proper statistical mechanical ensemble.  Because of
>     this, the temperature and pressure distributions it generates are
>     non-physical and the free energy surface of the simulation
>     trajectory is not correct.
> 
>     Let us know if you have any further questions.
> 
>     -Justin
> 
> 
>     Ravi Kappiyoor wrote:
> 
>         Sorry to bother you again, but I have some questions regarding
>         the MD simulation after getting the protein structure. As Joe
>         had suggested previously, I've used Modeler to create a
>         structure based on multiple templates (attached). I used 1KAP
>         and 1GO8 as they had better sequence identities (28 and 26%
>         respectively, as opposed to 21% and 19% for the other two
>         templates). I've been going through Justin's lysozyme and
>         umbrella sampling tutorials (except using the attached resilin
>         pdb instead of the pdb files that the tutorial files mention).
>         However, after going through both of them, I have a few
>         questions. In the lysozyme tutorial, after running an energy
>         minimization, an NVT minimization was done before an NPT
>         minimization, while the umbrella tutorial simply went straight
>         from energy minimization to NPT. I was wondering why NVT was
>         skipped in the umbrella tutorial. Also, in the coupling,
>         different coupling parameters were used between the two
>         simulations (the lysozyme used Parrinello-Rahman whereas the
>         umbrella used Brendsen coupling). A quick Google search didn't
>         seem to help me find the differences between the two, so I was
>         wondering if you had a link that would let me know the
>         difference between those two schemes. If not, I'll try and
>         search more thoroughly to see if I can find something. Sorry
>         once again to ask for so much help, but I've never worked with
>         MD before, so I'm trying to learn as much as I can before really
>         running any actual simulations.
> 
>         -- 
>         Thanks,
>         Ravi Kappiyoor
> 
> 
>     -- 
>     ========================================
> 
>     Justin A. Lemkul
>     Ph.D. Candidate
>     ICTAS Doctoral Scholar
>     MILES-IGERT Trainee
>     Department of Biochemistry
>     Virginia Tech
>     Blacksburg, VA
>     jalemkul[at]vt.edu <http://vt.edu> | (540) 231-9080
>     http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin
> 
>     ========================================
> 
> 
> 
> 
> -- 
> Thanks,
> Ravi Kappiyoor

-- 
========================================

Justin A. Lemkul
Ph.D. Candidate
ICTAS Doctoral Scholar
MILES-IGERT Trainee
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080
http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin

========================================