I am using Library MCS in Pipeline Pilot. We
have 2 parameters that we can change towards the outcome of clustering,
the method (exact, fast, very fast) and MCS size (default is 9). I
noticed very big clusters which actually embed two or more clusters
within each big cluster. Changing the MCS size or method does not
split these big clusters.
What is the parameter that I can use that will allow me to split these clusters.
With these two parameters you cannot directly split the clusters. Although, MCS size does influence the formation of clusters. If you set larger values for this parameter than the embedded clusters may not be merged: if their MCS is smaller than your predefined threshold then you end up with two clusters instead of one. But the threshold value can vary from clusters to cluster.
The embedded clusters you recognised by looking at the results aren't just artifacts, they are actual clusters. This is because LibraryMCS is a hierarchical clustering algorithm, but for various reasons the PP implementation does not reveal this hierarchy, it only outputs the top level clusters.
If you would like to get better insight in the hierarchy you may wish to try the LibraryMCS application
. We will also consider to integrate the hierarchical approach in PP.
hi i have installed LibMCS in a windos Xp system, i have to cluster my own set of compounds, so i collected some 130 structure sdf and united into one sdf and gave as a input, what happened is like the clustering happned for only 7% and showed as the clustering completed or terminated, so in result only 3 or 4 trees is formed remaining Structures were represedted as stright lines and no clustering was seen, so my question is whether all the clustering is over or is there any problem which i have to see in file format...i think so clustering is not done for rest of the structure.. so please help me out...
This might either be a problem in the program or in the input file. Does the input file open properly in MarvinView?