[gmx-users] Physical problems of comm removal

Christian Seifert cseifert at bph.ruhr-uni-bochum.de
Thu Feb 12 16:49:33 CET 2009 

Thanks for your ideas!

I will discuss the possible solutions with my supervisor.

To your PS:
This is a good hint, I changed many optimization flags to get a good
scaling but never changed the axis. I created the box with editconf
-princ. It automatically made z the longest axis and x the shortest.
Since many users use PME, it might be a good idea to change the behavior
of editconf -princ.

Again: Thank you very much for your help!
Christian. 

On Thu, 2009-02-12 at 16:28 +0100, Berk Hess wrote:
> Hi,
> 
> Another crude solution is to divide your protein into two groups along
> the long axis
> and use the pull code to keep the two components perpendicular to the
> axis
> of the vector between the COMs at zero.
> 
> PS: if you have PME your system will run faster when you make x the
> long axis.
> 
> Berk
> 
> > Date: Thu, 12 Feb 2009 16:21:29 +0100
> > From: cseifert at bph.ruhr-uni-bochum.de
> > To: gmx-users at gromacs.org
> > Subject: RE: [gmx-users] Physical problems of comm removal
> > 
> > Hi Berk,
> > 
> > the rhombic dodecahedron with 1771 nm^3 is still twice as big as my
> > stretched box.
> > 
> > This System scales quite well till 256 CPUs and I normally use this
> > amount of CPUs (I have to hurry up with my PhD thesis), so particle
> > decomposition is not an alternative.
> > 
> > Are there other alternatives?
> > 
> > Thanks so far!
> > Christian
> > 
> > On Thu, 2009-02-12 at 16:09 +0100, Berk Hess wrote:
> > > Hi,
> > > 
> > > The cleanest solution is using a rhombic dodecahedron:
> > > volume 13.58^3/sqrt(2) = 1771 nm^3
> > > (I don't know why people still consider truncated octahedrons).
> > > 
> > > You can try using comm-mode=angular with as comm group only the
> > > protein.
> > > This should not produce significant artifacts when your protein is
> > > large.
> > > However, this will not run with domain decomposition,
> > > only with particle decomposition (mdrun -pd), which scales worse,
> > > but probably not so bad with 4-8 cores.
> > > 
> > > Berk
> > > 
> > > > Date: Thu, 12 Feb 2009 16:00:30 +0100
> > > > From: cseifert at bph.ruhr-uni-bochum.de
> > > > To: gmx-users at gromacs.org
> > > > Subject: [gmx-users] Physical problems of comm removal
> > > > 
> > > > Hi users!
> > > > 
> > > > I simulate a stretched protein in TIP4P water in a cubic box
> > > > (6.83x9.60x13.58=890) with pbc. During a 50 ns simulation, my
> > > protein
> > > > must not rotate (otherwise, it will interact with itself). I
> found
> > > three
> > > > possibilities to avoid a self interaction: (1) a bigger box, (2)
> > > > restraints or (3) comm removal.
> > > > 
> > > > In details:
> > > > (1) bigger box:
> > > > I need at least a diameter of 13.58 nm in every direction to
> avoid a
> > > > self interaction (because of a possible rotation of the
> protein).
> > > For a
> > > > cubic box, this would be (13.58^3 =) 2504 nm^3. For a truncated
> > > > octahedron still 1928 nm^3, which would still be twice the box
> size
> > > I
> > > > use now. This is not suitable for me.
> > > > 
> > > > (2) restraints:
> > > > These are "long-time" runs, where I want to observe domain
> movement,
> > > so
> > > > restraints are not an alternative for my simulation.
> > > > 
> > > > (3) comm removal:
> > > > To use my stretched box, I have to use COMM-MODE = Angular. This
> > > mode
> > > > provokes artifacts, if I use it only on my protein and is
> therefore
> > > > forbidden since GMX4.0.3. But what happens, if I use it for my
> whole
> > > > system? My protein has a mass of 142800 a.m.u. and the solvent
> has a
> > > > mass of 998100 a.m.u. Let us assume, that the solvent movement
> > > reduces
> > > > itself, because its movement is undirected and the mass of the
> > > protein
> > > > is high enough to fall into account (which might not be true
> > > because:
> > > > 
> > > > v_{comm_removal}= 1 / (m_{solvent} + m_{protein}) \cdot
> > > > \sum_{i(solvent)}^n{(solvent)} ( m_i \cdot v_i) \cdot
> > > > \sum_{j(protein)}^o{(protein)} ( m_j \cdot v_j)
> > > > 
> > > > with \sum v_i=0 (because the movement of the solvent is
> undirected)
> > > the
> > > > equation would be:
> > > > 
> > > > v_{comm_removal}= 1 / (m_{solvent} + m_{protein}) \cdot 
> > > > \sum_{j(protein)}^o{(protein)} ( m_j \cdot v_j)
> > > > 
> > > > which weights the movement of the protein with its mass, put
> divides
> > > it
> > > > by the mass of protein+mass, what may result in a to small
> removal
> > > of
> > > > the comm.)
> > > > 
> > > > Beside this problem, do I have to expect that there might be
> other
> > > > problems with COMM-MODE = Angular, COMM-GRPS = System and a
> cubic
> > > box
> > > > with pbc?
> > > > 
> > > > 
> > > > 
> > > > -- 
> > > > M. Sc. Christian Seifert
> > > > Department of Biophysics
> > > > University of Bochum
> > > > ND 04/67
> > > > 44780 Bochum
> > > > Germany
> > > > Tel: +49 (0)234 32 28363
> > > > Fax: +49 (0)234 32 14626
> > > > E-Mail: cseifert at bph.rub.de
> > > > Web: http://www.bph.rub.de
> > > > 
> > > > _______________________________________________
> > > > gmx-users mailing list gmx-users at gromacs.org
> > > > http://www.gromacs.org/mailman/listinfo/gmx-users
> > > > Please search the archive at http://www.gromacs.org/search
> before
> > > posting!
> > > > Please don't post (un)subscribe requests to the list. Use the 
> > > > www interface or send it to gmx-users-request at gromacs.org.
> > > > Can't post? Read http://www.gromacs.org/mailing_lists/users.php
> > > 
> > > 
> > >
> ______________________________________________________________________
> > > Express yourself instantly with MSN Messenger! MSN Messenger
> > > _______________________________________________
> > > gmx-users mailing list gmx-users at gromacs.org
> > > http://www.gromacs.org/mailman/listinfo/gmx-users
> > > Please search the archive at http://www.gromacs.org/search before
> posting!
> > > Please don't post (un)subscribe requests to the list. Use the 
> > > www interface or send it to gmx-users-request at gromacs.org.
> > > Can't post? Read http://www.gromacs.org/mailing_lists/users.php
> > -- 
> > M. Sc. Christian Seifert
> > Department of Biophysics
> > University of Bochum
> > ND 04/67
> > 44780 Bochum
> > Germany
> > Tel: +49 (0)234 32 28363
> > Fax: +49 (0)234 32 14626
> > E-Mail: cseifert at bph.rub.de
> > Web: http://www.bph.rub.de
> > 
> > _______________________________________________
> > gmx-users mailing list gmx-users at gromacs.org
> > http://www.gromacs.org/mailman/listinfo/gmx-users
> > Please search the archive at http://www.gromacs.org/search before
> posting!
> > Please don't post (un)subscribe requests to the list. Use the 
> > www interface or send it to gmx-users-request at gromacs.org.
> > Can't post? Read http://www.gromacs.org/mailing_lists/users.php
> 
> 
> ______________________________________________________________________
> Express yourself instantly with MSN Messenger! MSN Messenger
> _______________________________________________
> gmx-users mailing list    gmx-users at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-users
> Please search the archive at http://www.gromacs.org/search before posting!
> Please don't post (un)subscribe requests to the list. Use the 
> www interface or send it to gmx-users-request at gromacs.org.
> Can't post? Read http://www.gromacs.org/mailing_lists/users.php
-- 
M. Sc. Christian Seifert
Department of Biophysics
University of Bochum
ND 04/67
44780 Bochum
Germany
Tel: +49 (0)234 32 28363
Fax: +49 (0)234 32 14626
E-Mail: cseifert at bph.rub.de
Web: http://www.bph.rub.de

More information about the gromacs.org_gmx-users mailing list