VMD-L Mailing List

From: viktor drobot (linux776_at_gmail.com)
Date: Sun Feb 23 2020 - 14:09:09 CST

Hi all. Recently I've been working a lot with AMBER Restart files and
used VMD for visualizing them. I've faced one small annoying bug with
VMD - it detects this type of input quite wrong. My AMBER version is 18
and VMD is 1.9.4a38.

Here is <https://ftp.desolve.ru/ftp/viktor/vmd_test.tar.gz>
(https://ftp.desolve.ru/ftp/viktor/vmd_test.tar.gz) my special test case
which consists of 5 files: topology (*system.prmtop*), coordinates with
timestamp (*crd.rst*), coordinates without timestamp (*crd_notime.rst*),
coordinates and velocities with timestamp (*crdvel.rst*), coordinates
and velocities without timestamp (*crdvel_notime.rst*). According to the
specs <https://ambermd.org/FileFormats.php#restart>
(https://ambermd.org/FileFormats.php#restart) timestamp field for plain
text restarts is optional. I think restarts without timestamp are
usually the very first input files for MD simulations right after tleap
or something similar and restarts with timestamp are just frames grepped
from trajectory.

I load parm + crd in VMD with the following filetypes: AMBER7 Parm
(topology) and AMBER7 Restart (coordinates). However I observe
inconsistent behavior for some restarts. Below I'll try to illustrate it
in detail.

1. Restart with coordinates and timestamp (*crd.rst*)

When loading this combination I can see the following in console:

rst7plugin) Title: Cpptraj Generated Restart

rst7plugin) This file contains velocity info.
rst7plugin) The Restartcrd has 85722 atoms.
Info) Using plugin rst7 for coordinates from file /tmp/1/crd.rst
rst7plugin) Importing velocities from restart file.
Info) Finished with coordinate file /tmp/1/crd.rst.

This file has no velocity info but has periodic box info. VMD tells that
it found velocities, however periodic box isn't found. Actually it
doesn't show anything and I see that total length of trajectory is 0.
So, only black screen without anything useful.

2. Restart with coordinates only (*crd_notime.rst*)

Console log:

rst7plugin) Title: Cpptraj Generated Restart

rst7plugin) This file has no velocity info.
rst7plugin) The Restartcrd has 85722 atoms.
rst7plugin) This restartcrd file has box info.
rst7plugin) Box Dimensions are 84.415970 100.727646 100.317093 90.000000
90.000000 90.000000
Info) Using plugin rst7 for coordinates from file /tmp/1/crd_notime.rst
Info) Finished with coordinate file /tmp/1/crd_notime.rst.

It tells that no velocities were found and reads in proper periodic box.
I can see my molecules on the screen so that combination works fine.

3. Restart with coordinates, velocities and timestamp (*crdvel.rst*)

Console output:

rst7plugin) Title: Cpptraj Generated Restart

rst7plugin) This file contains velocity info.
rst7plugin) The Restartcrd has 85722 atoms.
rst7plugin) This restartcrd file has box info.
rst7plugin) Box Dimensions are 84.415970 100.727646 100.317093 90.000000
90.000000 90.000000
Info) Using plugin rst7 for coordinates from file /tmp/1/crdvel.rst
rst7plugin) Importing velocities from restart file.
rst7plugin) Importing velocities from restart file.
Info) Finished with coordinate file /tmp/1/crdvel.rst.

It loads proper periodic box, detects velocities. I can see molecular
visualization. However it tells that velocities were read two times.
This is strange.

4. Restart with coordinates, velocities and without timestamp
(*crdvel_notime.rst*)

Info from console:

rst7plugin) Title: Cpptraj Generated Restart

rst7plugin) This file has no velocity info.
rst7plugin) The Restartcrd has 85722 atoms.
rst7plugin) This restartcrd file has box info.
rst7plugin) Box Dimensions are -0.243004 -0.109651 0.003363 0.591160
0.550960 -0.262683
Info) Using plugin rst7 for coordinates from file /tmp/1/crdvel_notime.rst
Info) Finished with coordinate file /tmp/1/crdvel_notime.rst.

It says that no velocities were found and also reads in incorrect
periodic box. However molecular visualization is shown on the screen.
Strange behavior too.

Seems like VMD relies on this timestamp field during parsing. However it
break things in some cases as shown above. I have already developed one
tool for working with AMBER restarts and it employs another approach.
According to the AMBER Restart specs we have the following structure of
plain text restart (/ denotes floor integer division and % denotes
modulo operator; so for 1 and 2 atoms in system we have only 1 line; for
3 and 4 we have 2 lines and so on):

HEADER
NATOM TIME
[NATOM / 2 + NATOM % 2] lines of X Y Z coordinates
[NATOM / 2+ NATOM % 2] lines of X Y Z velocities (if any)
PBC (if any)

So we just should count total number of lines in file and then subtract
2 (header and metadata). Let's denote (NATOM / 2 + NATOM % 2) as N and
total number of lines in file (not counting header and metadata) as T.
Then we have the following possibilities:

1. T = N. We have restart file with coordinates only and without
velocities and PBC.
2. T = N + 1. We have restart file with coordinates and PBC, but no
velocities are presented.
3. T = 2N. We have restart file with coordinates and velocities but
without PBC.
4. T = 2N + 1. We have restart file with coordinates, velocities and PBC.

Using this scheme VMD will open any modern AMBER Restarts correctly not
regarding timestamp at all (which is already optional). So what do you
think about it? Also why it reads velocities twice if they are presented?

Thank you!