[gmx-users] I NEED HELP USING G_CLUSTER

[Anton Feenstra]

Joseph Pichler wrote:
>     Dear Gromacs users,
> I am going to be running a MD simulation on a large system (NOS roughly
> 13,000 atoms and wish to cluster the simulation trajectory.  I wish to
> cluster the system into similar conformations.  I don't understand how the
> g_cluster command works.  If anyone has any information pertaining to the
> g_cluster command, please educate me.

It's fairly simple: choose an index group which you want to use for
comparing conformations (e.g. Ca atoms, or backbone or whatever you
feel is appropriate), and choose a clustering method. Depending on
which method you choose, there will also be a number of parameters
that you can play with. It depends a lot on the type of system and
type of dynamics it displays, which method would be best. Also the
optimal parameters are very much dependent on the situation.

Try it, have a look at some of the cluster central structures, and
also at some of the outliers. Compare different structures. Look
at how many clusters you have, and how large they are. Also look
at the RMSD within and between groups. A lot of this info is in
the cluster.log file, other info is in (optional) output files.

What I have often used is g_cluster in combination with essential
dynamics analysis (g_covar & g_anaeig), where you can make a
projection in 2d of your whole trajectory (the 'usual' essential
dynamics plots), but also of (several of) the clusters. That can
give you insight in the way the clusters are separated. In this
context, I also implemented (and used a lot) the 3d projection
option in g_anaeig (that produces a pdb file with dot representing
your trajectory, or cluster, conformations).

Hope this helps you on your way a bit!


-- 
Groetjes,

Anton
  _____________ _______________________________________________________
|             |                                                       |
|  _   _  ___,| K. Anton Feenstra                                     |
| / \ / \'| | | Dept. of Pharmacochem. - Vrije Universiteit Amsterdam |
|(   |   )| | | De Boelelaan 1083 - 1081 HV Amsterdam - Netherlands   |
| \_/ \_/ | | | Tel: +31 20 44 47608 - Fax: +31 20 44 47610           |
|             | Feenstra at chem.vu.nl - www.chem.vu.nl/~feenstra/       |
|             | "If You See Me Getting High, Knock Me Down"           |
|             | (Red Hot Chili Peppers)                               |
|_____________|_______________________________________________________|