Re: [AMBER] clustering MD results

From: Thomas Cheatham III <>
Date: Wed, 30 Mar 2011 15:28:39 -0600 (Mountain Daylight Time)

> I've tried to analyze MD results with clustering. I found manual of ptraj program with following example that produces 5 clusters using a RMS metric comparing the CA atoms:
> trajin traj.1.gz
> trajin traj.2.gz
> cluster out testcluster representative pdb \
> average pdb means clusters 5 rms .CA
> Now, I won't specify the number of the clusters, but woule like to produce clusters based on the RMSD difference at cutoff = 2 angstrom.
> How should I modify the script?

The AmberTools ptraj manual says:

  [clusters n | epsilon critical-distance]

This means either say "clusters n" or "epsilon n", so replace "clusters 5"
with "epsilon 2.0" noting that later it says this only works with
averagelinkage, linkage, complete, edge, centripetal, centripetalcomplete
or hierarchical clustering. Averagelinkage is probably a good alternative
to means.


AMBER mailing list
Received on Wed Mar 30 2011 - 15:00:02 PDT
Custom Search