Quite frequently LibMCS would present an incomplete ring systems as the maximal common substructure for a set of molecules. For my purposes this doe snot help as I am interested in identifying substructures for perfroming SAR-type analysis. Is there a way to force LibMCS to only identify complete ring systems as maximal common substructures.
I'm afraid this functionality is not yet available in LibMCS. Something similar is provided by the Bemis-Murcko frameworks, that can be used to group compounds (I would not call that clustering, but it's related to clustering). You may wish to give it a try online at http://discoverygroup.chemaxon.com/MGSandbox/jkdemo.jsp. Do you think that can be relevant for you?