[gmx-users] trjconv -pbc cluster with rhombic dodecahedron box

[chris.neale at utoronto.ca]

Hi Daniel,

1a. I am confused about what is your vesicle. Is it everything that  
you are showing in these images? Or perhaps it is only the cyan thing  
that I can see sliced through in:  
http://img109.imageshack.us/img109/8227/trajenddat.jpg

1b. I suspect that the cyan ring encircled by green in the above  
mentioned image is actually the beginning of a pbc-merger event of  
your vesicle with itself. If this is true, then definitely refer to  
point 2 below. This would be easier to see if you create an image of  
the trajenddat.jpg in which you display a number of periodic images.

2. In addition to what Tsjerk has mentioned, your might need a larger  
simulation box in order to simulate your vesicle properly. If it is  
actually interacting with itself then you may be sampling a state that  
is actually uninteresting to you. At the very least, it is more  
complex than it needs to be.

Regarding the rebuilding procedure that Tsjerk has mentioned, we are  
in the process of coding this functionality into gromacs. I believe  
that Berk also has a slightly different solution here. Nevertheless,  
no amount of pbc finessing is going to change the fact that you are  
simulating a state that appears to be different than what you are  
interested in.

Chris.

-- original message --

Hi Daniel,

The problem is likely that your vesicle is interacting with itself
over the periodic boundaries. There are regions where there is no
solvent inbetween. That means that lipids can go over from one image
to the other by diffusion, which will not be compensated by using -pbc
nojump. You seem to be out of luck there.
The solution to solve this problem would be to rebuild the vesicle by
minimizing the sum of squared deviations from the center using shifts
over the lattice. But that is not implemented.

Cheers,

Tsjerk


On Tue, Dec 1, 2009 at 2:55 PM, Daniel Parton
<daniel.parton at bioch.ox.ac.uk> wrote:
> Hi,
>
> I've tried using trjconv -pbc nojump with a .tpr created from the first
> frame of the trajectory (clustered so the vesicle should be whole). The
> output trajectory starts with a complete vesicle, but small numbers of lipid
> particles progressively move to places outside of the box. By the end of the
> trajectory there appear to be 8 groups of particles outside the box, which
> have the same overall spherical shape as the vesicle, but are comprised of
> only a few hundred particles. I have uploaded some pictures here:
> http://img109.imageshack.us/g/trajenddat.jpg/
>
> traj-start is the first frame of the unprocessed trajectory:
> http://img40.imageshack.us/img40/2181/trajstartdat.jpg
>
> traj-end is the last frame of the unprocessed trajectory:
> http://img109.imageshack.us/img109/8227/trajenddat.jpg
>
> nojump-start is the first frame of the trajectory following trjconv -pbc
> nojump, as described above. The vesicle starts whole:
> http://img138.imageshack.us/img138/4531/nojumpstartdat.jpg
>
> nojump-end is the last frame of the trajectory following trjconv -pbc
> nojump. Note that some particles have moved to being outside the box,
> seeming to take up space resembling 8 copies of the vesicle system.
> http://img689.imageshack.us/img689/2821/nojumpenddat.jpg
>
> 0ns-cluster is the first frame of the trajectory, clustered using trjconv
> -pbc cluster. The pbc option may have gone a bit wrong? Particles which
> should be very close to the box boundary seem to have been moved to other
> places in or outside the box. Other parts which lie some way inside or
> outside the box appear (correctly) as part of the vesicle:
> http://img222.imageshack.us/img222/4268/0nsclusterdat.jpg
>
> So I am wondering the problem lies with the 0ns-cluster system. As some
> particles are not part of the vesicle in this representation, this could
> explain why some particles are not being counted as part of the vesicle by
> trjconv -pbc nojump. Does this seem likely, and has anyone seen similar
> behaviour to that seen in 0ns-cluster, where trjconv -pbc cluster has been
> used with a rhombic dodecahedron box?
>
> Many thanks for all of your help so far,
>
> Daniel