[gmx-users] clustering and representative structure

[Nikos Sgourakis]

Hello,

The most cited method on clustering is probably k-means. Although not 
included in the current gromacs version, I have implemented it within 
the gromacs gmx_cluster code and you can find it in the archives.
All of these methods (excluding principal components) can offer central 
actual structures that are representative of your sample. Since there is 
no "best" clustering but the output depends strongly on the distribution 
of structures in your dataset, no method is guaranteed to find the most 
representative structures. However, you can obtain a consensus of 
structural features, through a comparison of the central structures from 
different methods.
A very good reference on clustering is the book by J.A. Hartigan 
"Clustering Algorithms".

Best regards,
Nikos Sgourakis


Carlos Javier Nuñez Aguero wrote:
> Hi all,
> How are you?
>
> In Gromacs, the clustering methods are:
>
> o Full linkage
> o Jarvis Patrick
> o Monte Carlo
> o Diagonalization, and
> o Gromos
>
> ---- what is the most cited method?
> ---- what option generate "the most representative structure"?
>        all the strategies? why?
>       (not the average geometry)
> ---- where I can read more (or where I can find specific information) 
> about of
>       every methodology of clustering?
>       links, papers, etc? (moreover of the manual or online or html 
> help).
>
> Thanks,
> Regards,
> Carlos Javier
>
> _______________________________________________
> gmx-users mailing list    gmx-users at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-users
> Please don't post (un)subscribe requests to the list. Use the www 
> interface or send it to gmx-users-request at gromacs.org.
> Can't post? Read http://www.gromacs.org/mailing_lists/users.php
>
>
>
-- 
___________________________
Nikolaos G. Sgourakis, MSc
Center for Biotechnology and Interdisciplinary studies
Troy, NY 12180
www.rpi.edu/~sgourn