Re: [AMBER] Self-Diffusion coefficient

From: Bruno Rodrigues <bbrodrigues.gmail.com>
Date: Tue, 20 Sep 2011 15:13:53 -0300

On Mon, Sep 19, 2011 at 7:50 PM, Thomas Cheatham III <tec3.utah.edu> wrote:

>
> > I was reading the manual to find a way to put a mask for the water
> molecules
> > in order to compute only the ones that are closer to the DNA, but none of
> > them worked.
> >
> > trajin 1D20_wat_tip3pfNVE310.x 1 99999
> > diffusion (:1-20<:5.0)&:WAT 0.1 average 310_within5a
> > diffusion :WAT&(:1-20>:8.0) 0.1 average 310_without8a
> >
> > However, it seems that the mask is not selecting the expected atoms, as
> can
> > be seem from the output:
> >
> > PTRAJ: diffusion (:1-20<:1.0)&:WAT 0.1 average 310_within5a
> > PTRAJ: diffusion :WAT&(:1-20>:8.0) 0.1 average 310_without8a
>
> As currently implemented, I believe this selection is based on the first
> frame and therefore the mask is not dynamic (i.e. it is not updated or
> changed for each frame. [Dynamic updates of the mask would not only be
> very costly, but tricky for averaging as the #atoms could change for each
> frame; it would likely require modifying MANY of the ptraj action routines
> to understand the dynamic behavior].
>
> This fixed list explains why similar results are obtained since as the
> water diffuses away from the DNA, it approaches the bulk values
> independent of the selection (given sufficient time).
>

Well, this now make sense for me. I didn't realize the static definition of
the mask. I found, however, this code ( I don't know if it has been tested
by the Amber developers, but I'm trying to run some tests, but it is very
slow and from now no results were output).
http://www.cse.scitech.ac.uk/cbg/software/ptraj/


>
> If the mask *did* dynamically select the subset of atoms at each frame,
> i.e. for a mobile "layer" since "free" water exchanges rapidly in and out
> of the shell at the surface and may rarely move at most bound sites, the
> medium time scale diffusion statistics would likely be extremely noisy.
> As diffusion is a "bulk" property, I would be concerned about the
> statistics in general for a subset of the system; an easy way to visualize
> this is to compare the estimation of diffusion for a single water (or
> single ion) compared to all waters. Very noisy unless you have very long
> trajectories.
>
> Certainly to do this diffusion as a function of distance will require your
> own code (although perhaps it has been implemented by others and I simply
> do not know this). To do this in ptraj, a starting point could be the
> diffusion code which would have to be (copied into a new routine and) made
> smart enough to keep track of each individual water (i.e. all waters)
> and/or whether it was in the shell (or while I was at it, since you are
> tracking all waters, might as well do all shells at the same time at the
> same cost). The tricky part would be figuring out the averaging and of
> course getting results that are not inundated by noise. For DNA, I would
> likely want this result as a function of distance from the helical axis
> which complicates things further (i.e. it wouldn't be spherically
> symmetric and may be different in grooves/backbone). Perhaps you could
> get away with monitoring distance from a particular functional group only
> at short range distances.
>
> I think this can be implemented in VMD by using the tcl language. I will
try to first reproduce the bulk msd and then come to a more refined code
where I can play with the masks.

>
> To better understand positional dependent dynamics and Before I went in
> and modified the code as a starting point I would look at the work by
> Furse/Corcelli (multiple papers) and/or Sen/MA Berg (JACS, 2009, 131,
> 1724). Perusing PubMed, I would also take a look at Lin/Goddard JPCB
> 2005, 109, 8663 and/or Korolev/Nordenskiold NAR 2003, 31, 5971 and
> earlier. There is also older work on proximal density functions by
> Beveridge and Mezei Methods Enzym. 1986, 127, 21. These may or may not do
> what you want or provide guidance. I would also look at other analysis
> tools (perhaps in CHARMM and/or GROMACS/Gromos) to see if anything was
> implemented.
>
> But I would say, scientifically speaking, that the behavior of the
diffusion coefficient from the DNA surface is already known since a long
time ago (following the work of Montgomery Pettitt). Anyway I observed an
overall decrease of the diffusion coefficient wrt the bulk water, and this
is enough for me.



> --tec3
>
> p.s. no one ever said research was easy nor that ptraj could do everything
> or even most things useful; the analysis tools are still evolving and
> contributions are welcome...
>
> That's correct. Sometimes I feel we get a little bit stuck by the code and
its capabilities, and forget that the research itself has no boundaries and
must overcome the available tools.


> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>



-- 
-- 
Bruno Barbosa Rodrigues
PhD Student - Physics Department
Universidade Federal de Minas Gerais - UFMG
Belo Horizonte - Brazil
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Sep 20 2011 - 11:30:03 PDT
Custom Search