Ligand/inhibitor density in active site?

[olibclarke]

Have you found using class similarity ~=0 gives better/different results to enabling force hard classification in such cases?