Issue with rhombic dodecahedron / CHARMM compatibility

From: Dan Bolintineanu (bolintin_at_msi.umn.edu)
Date: Mon Dec 04 2006 - 17:03:21 CST

I'm trying to run some solvated MD simulations using a rhombic
dodecahedron (RHDO) box shape. I want to do the model building in
CHARMM, then use NAMD for the production runs, then use CHARMM for
analysis. NAMD claims to be fully compatible with the CHARMM
trajectory format, but as it turns out, this doesn't seem to hold for
RHDO simulations. After successfully building and minimizing the model
in CHARMM, I try to load it into NAMD (after using the crd2pdb
conversion utility).
In NAMD, the van der Waals energy is through the roof, due to bad
image contacts - I know this because using a much larger box
definition in NAMD results in a reasonable energy.
I tried COOR CONV ALIGNed SYMMetric before exporting the CHARMM crd
file, but no avail.
If I then re-build the model in NAMD, and produce a trajectory, I
cannot read it in CHARMM and perform any image operations. I know
there is a compatibility issue with the way unit cell angles (alpha,
beta, gamma) are stored, but I think there is also a compatibility
issue with the orientations of the unit cell.

I got the CellBasisVectors corresponding to the RHDO shape from the
GROMACS manual, which lists two orientations for the RHDO shape: one
which has a square cross-section with the XY plane, the other with a
hexagonal cross-section. I've tried every combination of COOR CONV
ALIGN/SYMM and the two different cell basis vector configurations. I
know that the cell basis configurations DO represent an RHDO, since
extensive minimization and dynamics in NAMD does result in an
RHDO-shaped water box (when viewed in VMD).

Furthermore, I have tried several of the available TCL scripts to
unwrap a NAMD trajectory produced with the above setup, and they all
yield ridiculous output. Has anybody been able to successfully port
RHDO-based simulations between charmm and NAMD, or at least
wrap/unwrap an RHDO simulation produced by NAMD?

Any advice on this would me much appreciated.

Dan Bolintineanu

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 05:19:55 CST