I think others have made the important points here, but I'll just try to
clarify a bit. There is a distinction between a molecular dynamic
simulation and an X-ray structure. In the MD simulation, we track each
individual molecule on its journey throughout the simulation cell. In an
X-ray structure, we track the average density of matter, not knowing which
molecule it came from or even what type of molecule it came from. All of
the X-ray structures you see have been refined by making inferences about
the known composition of the crystal and the known masses of the elements
and basic sub-arrangements in them.
Therefore, if you do an RMS fit of your simulation around your protein and
then compute the average positions of all residues, you might expect to be
able to see which components of the protein itself are relatively
stationary and which parts are floppy (the RMS fluctuations of each atom
are what you'd look at: if they're small, then they're more or less
stationary and the local, average structure will be something
recognizable--if they're big the average structure will probably look
weird). If you want to compute an average density, you will need to use
some of cpptraj's histogram tools (think in terms of three-dimensional
histograms, making bins with counts of water and lipid molecules relative
to the average structure of the protein). Another analysis that might be
of interest to you is clustering. Try out the "cluster" command and see
what you can do.
Dave
On Sat, Mar 26, 2016 at 6:14 PM, Chris Moth <cmoth08.gmail.com> wrote:
> It is of course possible to compare the backbone RMSD of all snapshots to
> the average, and then excerpt the closest snapshot, which will certainly
> have reasonable solvent and counter-ion positions. But, interpreting the
> average structure, or a close snapshot, as "telling us something", would
> most likely be a case of over-interpretation. If the MD trajectory
> ensemble is "telling you something" - backed by the statistics, _and_ there
> is a snapshot that illustrates that point, then OK.. that might have
> application...... within the caveat that it still only a snapshot.
>
> On Sat, Mar 26, 2016 at 3:45 PM, Bill Ross <ross.cgl.ucsf.edu> wrote:
>
> > Imagine a moiety that twists and flaps back and forth with respect to
> > the rest of the system. A raw average might be an unchemical jumble in
> > the middle. My impression is that this situation is less likely in
> > crystals.
> >
> > Quick experiment - run vacuum dynamics and average to see what I mean.
> >
> > Bill
> >
> > On 3/26/16 12:45 PM, Michael Shokhen wrote:
> > > Dear Jason,
> > >
> > > Thank you for your response.
> > > You wrote: "It is generally a mistake to think that an average
> structure
> > will be
> > > physically realistic."
> > > What is a crystal structure of protein molecule identified by x-ray?
> > > Isn't it an average structure over all protein molecules in real
> > physical system?
> > > Following your statement what single frame on the MD trajectory should
> > be used
> > > as a resulting computational structure simulating a protein structure
> in
> > a real physical system?
> > >
> > > Regards,
> > > Michael
> > >
> > > *****************************
> > > Michael Shokhen, PhD
> > > Associate Professor
> > > Department of Chemistry
> > > Bar Ilan University,
> > > Ramat Gan, 52900
> > > Israel
> > > email: shokhen.mail.biu.ac.il
> > >
> > > ________________________________________
> > > From: Jason Swails <jason.swails.gmail.com>
> > > Sent: Saturday, March 26, 2016 10:18 PM
> > > To: AMBER Mailing List
> > > Subject: Re: [AMBER] membrane protein problem
> > >
> > > On Sat, Mar 26, 2016 at 2:35 PM, Michael Shokhen <
> > michael.shokhen.biu.ac.il>
> > > wrote:
> > >
> > >> Dear AMBER experts.
> > >>
> > >>
> > >> Applying AMBE14 and AmberTools15, I have simulated a protein in
> membrane
> > >> by MD in periodic cell.
> > >>
> > >> The next step is calculation of the average structure from
> > >> the fragment of the production trajectory that satisfies a desired
> > protein
> > >> backbone RMSD condition.
> > >>
> > >> Finally, to obtain a representative protein structure the latter step
> > must
> > >> be minimization.
> > >>
> > >>
> > >> The adequate minimization of a membarne protein can be conducted in
> > >> membrane environment only. That is why I have accounted in the
> averaging
> > >> procedure
> > >>
> > >> the whole soup in the periodic cell: protein, membrane, water, and
> > counter
> > >> ions.
> > >>
> > >> In this purpose I have applied the following cpptraj script:
> > >>
> > >>
> > >> parm ../*.prmtop
> > >>
> > >> trajin prod4.mdcrd 3000 10000
> > >>
> > >> center :1-9308 mass origin
> > >>
> > >> autoimage origin
> > >>
> > >> rms first .C,CA,N
> > >>
> > >> average average_prod4_3000-10000.pdb PDB
> > >>
> > >> run
> > >>
> > >> quit
> > >>
> > >>
> > >>
> > >> The problem I faced is that despite the final pdb file contains also
> > >> solvent water molecules and counter ions their coordinates are
> > corrupted.
> > >>
> > >> Water molecules and counter ions are collapsed
> > >>
> > >> in a very small condensed area in the center of periodic cell
> > >> like in a black hole.
> > >>
> > > This is exactly what I would expect to happen in a converged
> simulation.
> > > Picture a droplet of water that is suspended in zero-G (free-fall) for
> a
> > > long time. Each molecule is constantly moving, and the ready exchange
> of
> > > water (due to typical diffusion) means that each water molecule spends
> > > equal time at every "location" within the droplet as long as it has
> > enough
> > > time to diffuse everywhere. If its probability density in the droplet
> is
> > > uniform, its *average* position is the center of the droplet. And this
> > > holds true for *every* water molecule.
> > >
> > > It is generally a mistake to think that an average structure will be
> > > physically realistic.
> > >
> > > HTH,
> > > Jason
> > >
> > > --
> > > Jason M. Swails
> > > _______________________________________________
> > > AMBER mailing list
> > > AMBER.ambermd.org
> > > http://lists.ambermd.org/mailman/listinfo/amber
> > > _______________________________________________
> > > AMBER mailing list
> > > AMBER.ambermd.org
> > > http://lists.ambermd.org/mailman/listinfo/amber
> > >
> >
> >
> > _______________________________________________
> > AMBER mailing list
> > AMBER.ambermd.org
> > http://lists.ambermd.org/mailman/listinfo/amber
> >
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sat Mar 26 2016 - 16:00:04 PDT
