> 1.How can I know the time scale of each cluster?
Not directly... However there is a file produced, the *.txt file that has
information for each cluster on the time course (where each character is a
snapshot and X means occupied, dot means not). A summary of this is
printed at the bottom of the file:
#Clustering: divide 16334 points into 5 clusters
#Cluster 0: has 23 points, occurence 0.001
#Cluster 1: has 1011 points, occurence 0.062
#Cluster 2: has 10984 points, occurence 0.672
#Cluster 3: has 1 points, occurence 0.000
#Cluster 4: has 4315 points, occurence 0.264
#Cluster 0 .
#Cluster 1 99241 11 .
#Cluster 2 .758XX88XXX99XXXXXXXXXXXXXXXXXXXX9X8.
#Cluster 3 .
#Cluster 4 .. 19XXXXXXXXXXXX
No character means nothing over that time interval with higher numbers and
eventually X meaning higher occupancy/total occupancy over that interval.
>From the plot above cluster 0 was very short lived, moved to cluster 1,
then 2 back to 1 back to 2 to a short lived 3 and then cluster 4.
Another way to visualize "clustering" is with the 2D-RMSd plot.
> 2. Here is the output file of the cluster analysis.
>
> #Desired Clusters: 5
> #Cluster 0 has average-distance-to-centroid 2.837542
> #Cluster 1 has average-distance-to-centroid 2.250712
> #Cluster 2 has average-distance-to-centroid 1.753686
> #Cluster 3 has average-distance-to-centroid 2.600417
> #Cluster 4 has average-distance-to-centroid 2.933015
> #DBI: 1.18191
> #pSF: 4607.87695
> #SSR/SST: 0.64839
>
>
> What is the centroid? Is it the average structure?
The centroid is the cluster center.
> And in the cluster pdb, I find that the cluster0 is very similar to
> the initial structure and the cluster 4 is very different with the
> initial structure.
>
> But in the output it shows their distance is 2.8 and 2.9. They are
> similar. Their structures are very different but the distances are
> very similar. Why?
That distance reported above is the average distance of each member of the
cluster to its cluster centroid. So, although cluster 0 and 4 are far
apart, the structures in each cluster are similar.
--tec3
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sun Nov 06 2011 - 15:00:02 PST
