[gmx-users] Re: g_cluster settings

[Ran Friedman]

Marc - Note that you used the linkage method which maps a structure to a
cluster if its distance to any structure already in the cluster is is
less than the cutoff. If you think it's not representative of your
system try another clustering method, but note that you may need a
longer trajectory as well.

Vitali - g_clustersize is for clustering molecules (can be useful e.g.,
for the formation of micelles) whereas g_cluster is for clustering the
trajectories.

Ran

Vitaly Chaban wrote:
> Dear Marc:
>
> Please look towards g_clustzise (instead of g_cluster). We tried to
> use g_cluster some years ago and eventually switched to g_clustzise -
> I do not remember why but the latter was more successful. If you want
> to monitor the number of clusters in the system, g_clustzise can
> certainly do it.
>
> Dr. Vitaly Chaban
>
>
>
>   
>>     I am trying to perform a cluster analysis on a 4 ns trajectory (4000 frame
>> trajectory), but I am not getting any more than 1 cluster until I get down to a
>> cutoff of 0.05 nm. This seems like an awfully small number, so I checked my rmsd
>> distribution and found that 44% of my rmsds are at 0.1 nm or greater. My
>> distribution looks like the following:
>>
>> And my command line is as follows:
>>
>> g_cluster -f md.trr -s md.tpr -n index.ndx -sz clustersize1.xvg -clid
>> clusterid1.xvg -cutoff .10   -cl clusters1.pdb -b 1 -e 4001