[gmx-users] pdb2gmx: Amber trajectory conversion

Mark Abraham Mark.Abraham at anu.edu.au
Wed Jul 19 04:18:41 CEST 2006

Dallas B. Warren wrote:
> The script that deals with converting trajectories is trjconv.  Have a
> look at the documentation, that will show you the formats it can handle,
> no idea what Amber format is, but as long as you can get it into one of
> the ones it can use should be fine. 

trjconv can't handle the usual Amber DCD format

> Plus, with g_cluster I suspect you
> will require a topology file to be able to do the analysis, which will
> be your major sticking point.

Actually it needs a structure file - in man g_cluster the FILES section 
indicates that the -s option needs one of many formats that imply 
structure and masses, and a full .tpr is thus not required.

> Note that pdb2gmx converts pdb coordinate files to generate a topology
> file (and the residues/molecules have to be known by the database
> otherwise you have to make by hand etc).

Here's the text of an email I sent to a group member who was grappling 
with this general AMBER conversion issue just last week...



Hi Sue,

Issue 1 - using gromacs cluster tools with amber trajectories.

Converting the trajectory is easy - load the structure into VMD, import 
the amber trajectory and choose Save Coordinates from the File menu and 
choose the "trr" file format. There's an ugly Plan B if that doesn't 
work, but I won't describe it now.

You'll also need any structure file that will supply atom names and 
stuff, and use the gromacs tool make_ndx to make a index file that 
basically describes the subset of atoms you actually want to do the 
analysis on - i.e. the atoms for the RMSD.

Those three files will be enough to run g_cluster with, so actually it 
should be very painless for you. The most annoying bit would be 
transferring the trajectory files to and from ac to use vmd on them - 
hang on, they have vmd 1.8.2 available on ac ("module load vmd") if you 
have X-Windows available (e.g. markov or valera console) then you can 
even do that conversion remotely.

I'd suggest getting the gromacs manual from their website 
http://www.gromacs.org/. The man pages for their tools are included 
there, but "man make_ndx" and "man g_cluster" will give you most of what 
you need to know and you're welcome to ask me for clarification. Note 
that on those manpages the SYNOPSIS section implies particular filename 
extensions - actually there's normally a range of possible formats that 
are described fully in the FILES section for each option. In particular, 
I don't believe you'll need to generate a .tpr file - a .pdb produced by 
ambpdb ought to suffice.

Issue 2 - clustering and VMD
If you look here http://www.ks.uiuc.edu/Research/vmd/plugins/ there is a 
clustering-result visualisation package available for VMD but it 
requires the use of two clustering codes I know nothing about, at least 
one of which will cost money - XCluster.

I think the gromacs solution is the way to go initially. I am not sure 
how to generate views of simultaneous superpositions in VMD, but the 
gromacs tool trjconv makes it easy to extract a series of snapshots from 
the trajectory that correspond to a cluster into (say) the same PDB file 
and now it should only be a matter of massaging that to get VMD to show 
them superposed.

Oh, and there is also KClust http://mmtsb.scripps.edu/doc/kclust.html 
which might be easiest of all, but you're on your own here.


More information about the gromacs.org_gmx-users mailing list