Re: [AMBER] Average Conformations from trajectory

From: Daniel Roe <>
Date: Thu, 5 Sep 2013 08:11:06 -0600


On Thu, Sep 5, 2013 at 7:33 AM, Covington, Cody Lance
<> wrote:
> I have some simulations of a small molecule in water and in a micellar aggregate. I was wanting to analyze the conformations present in both trajectories for comparison. I would like an output of the most common conformations with the percentage of time that the structures were occupied.

It sounds like what you may want is to perform cluster analysis, which
groups structures together based on some distance metric. Cluster
analysis can be performed with either ptraj (AmberTools 13 manual
section 9.3) or cpptraj (section 8.13.1). Cpptraj allows you to
explicitly cluster and compare results from two trajectories in one
run via the 'summaryhalf' and 'splitframe' keywords. For example, say
you have two trajectories, which has 500 frames and
which has 1000 frames, the following cpptraj input will perform
hierarchical agglomerative clustering using coordinate RMSD as a

cluster hieragglo epsilon 1.0 clusters 5 averagelinkage rms <mask>
summary summary.dat info info.dat summaryhalf half.dat splitframe 501
repout rep.pdb repfmt pdb

Here <mask> would correspond to the molecule you want to cluster on.
You will probably have to experiment a little with the 'epsilon' and
'clusters' parameters to obtain optimal results. You could also try
the 'dbscan' algorithm, or try clustering on other data sets
(distances, angles, dihedrals etc - see the manual for more details).
If you have a lot of frames (> 10000 total) you will probably want to
use the 'sieve' option. It's probably best to just do some inital
clustering on a small subset of frames first anyway just to become
familiar with how it works. Note that the pairwise distance calc. is
parallelized in the OpenMP version of cpptraj.

Ptraj has more clustering algorithms available, but has slightly
different syntax and does not have the builtin 'summaryhalf' analysis.
A good reference on clustering using ptraj (and at least hierarchical
agglomerative clustering in cpptraj) is:

Shao, J.; Tanner, S.W.; Thompson, N.; Cheatham, T.E. III. J. Chem.
Theory Comput., 2007, 3, 23122334.

Hope this helps,


Daniel R. Roe, PhD
Department of Medicinal Chemistry
University of Utah
30 South 2000 East, Room 201
Salt Lake City, UT 84112-5820
(801) 587-9652
(801) 585-9119 (Fax)
AMBER mailing list
Received on Thu Sep 05 2013 - 07:30:02 PDT
Custom Search