[gmx-users] Analysis programs and PBC.

[Daniela S. Mueller]

Dear Maxim, dear David, 

On Sun, 2006-02-26 at 22:34 +0100, David van der Spoel wrote:
> Maxim Fedorov wrote:
> > Dear all,
> > 
> > I have been successfully using Gromacs for a couple years without 
> > bothering you -
> > manual + reading the mailing list was completely enough.
> > 
> > But … this is the second mail for this afternoon.
> > 
> > The question is: are there any analysis programs in GROMACS package 
> > which are sensitive to Periodic Boundary Conditions or they remove the 
> > periodicity automatically?
> > 
> > Now I explain in more details what I mean:
> > 
> > Let’s suppose that I have a trajectory file after some MD run with 
> > PBC. No shuffling, sorting and other things like that – only PBC.
> > 
> > Then I am going to analyze the trajectory with g_rama, g_hbond, g_rdf, 
> > g_gyrate and, very probably, something else.
> > Should I remove the periodicity first or I will be able to analyze the 
> > trajectory without bothering myself about that issue?
> > I have already tried to analyze the PBC runs with and without removal 
> > of periodicity– the results are the same. 
> > But may be this is only due to the fact that my molecule didn’t move 
> > too much during the run and was placed as a whole in the periodic 
> > cell? And, in a general case I have to remove the periodicity to be 
> > 100% in the result of analysis.
> > 
> > Sorry, for this, probably, silly question, but I am starting several 
> > series of simulations - I will be not physically able to check every 
> > time several hundreds of trajectories with or without periodicity 
> > removal
> > (even with modern scripting languages a human being should take a look 
> > on the final results, isn't it :-)).
> > 
> 
> In all programs were it is needed PBC is taken care of. I have recently 
> built in flags into some of the programs that allow you to control the 
> removal of periodic boundary conditions (i.e. putting whole molecules in 
> the box prior to analysis). The reason for this is that we're doing 
> extensive simulations in vacuo, and putting the molecules back in the 
> box breaks the analysis. In solvent the default behavior is fine.
> 
> Bottom line is that PBC is treated properly as far as I know. But you 
> should always check your output :).
> 

If the molecule you want to analyse is a single molecule, the jumping of
box boundaries does not matter and the analysis tools take that into
account. But if your molecule is a complex of two or more separate
subunits, you might want to remove jumps over the boundaries to
reassemble the complex again, e.g. to measure the minimum distance to
the periodic images (and not to the subunit, which should be very
nearby). 

What I actually found with trjconv -pbc nojump was, that two conditions
apply if you want to remove jumps: 1) the trajectory has to start at the
initial frame and be continuous; 2) the tpr-file should correspond to
the initial conformation. If these conditions weren't fulfilled, one of
the subunits was rotated and the complex was reassembled incorrectly -
although the box was rectangular. 
This incorrect rotation seems to be a weird bug, probably the
referencing to the tpr-structure is a special case-treatment and not
robust to various setups...? 

Success and best regards, 
Daniela


> > With kindest regards,
> > 
> > Maxim.
> > 
> > Maxim V. Fedorov
> > Research Associate.
> > 

-- 

Daniela S. Mueller

biologist (Diplom, German degree)
______________________________________________________________________

- Molecular Dynamics Group, UQ -

Address:
School of Molecular and Microbial Sciences (SMMS)
Chemistry Building (#68)
University of Queensland
Qld 4072, Brisbane
Australia

Phone: +61-7-33653732

Website: http://ilc00f.facbacs.uq.edu.au/SMMS/a_mark/Front.htm

**********************************************************************

- MD group, RuG -

Address:
Molecular Dynamics
Dept. of Biophysical Chemistry
University of Groningen
Nijenborgh 4
9747 AG Groningen
The Netherlands

Website: http://www.rug.nl/gbb/md
______________________________________________________________________