From: Bob Johnson (
Date: Thu Mar 15 2007 - 16:08:50 CDT

Hello everyone,
I'm trying to visualize/analyze the results of a replica exchange MD (REMD)
trajectory of a ssDNA molecule. The trajectory was calculated with Gromacs
which only writes the coordinates of atoms that are within the primary box. As
expected, during the trajectory, portions of the ssDNA that move into the next
periodic box are wrapped back into the primary box. As a result, the ssDNA
molecule appears to be broken and VMD draws long bonds from one end of the box
to the other.

Normally, I can correct for this using either the PBCtools unwrap plugin or the
Gromacs trjconv tool with the -pbc nojump option. However, the nature of the
REMD trajectory causes both of these options to fail.

I believe that the way both of these tools work is that they detect "jumps"
across the box and then correct for them by shifting the atoms back. These
"jumps" are simply large displacements between successive frames. However, in
REMD there are large displacements between successive frames due to the
instantaenous coordinate swaps. Thus, attempting to use the aforementioned
tools on a REMD trajectory leads to detection of erroneous "jumps".

I had the idea of writing a script that would remove PBC wrapping based on bond
length. If the bond in VMD gets larger than a certain amount, the script would
detect this as a jump. Then, the script would translate the appropriate atoms
in the appropriate direction to make the molecule whole again. I'm assuming VMD
stores information about what atoms share a "chemical bond" (i.e. are connected
by a line). Is there a way to access this information? My idea is still somewhat
rough so any suggestions or insights on this would be appreciated!
Bob Johnson