I have one remark and one request (if possible) about scaffolds and frameworks generation with the indigo nodes.
With the scaffold generator node I obtain in results scaffolds with a C=O group as linker between 2 cycles. Same when the C=O is in a cycle, this group keep in place during the scaffold generation. In my meaning the definition of Murcko scaffold imply that these carbonyl group should be supressed.
After more verifications it is the same at least for S=O and SO2 linkers.
In meaning we should obtain:
Molecule --> Scaffold
Ph-CO-Ph --> Ph-C-Ph
Ph-SO-Ph --> Ph-S-Ph
Ph-SO2-Ph --> Ph-S-Ph
For the framework study, you provide a very interesting and usefull node to replace all the atom by an other (here C). But is-there a solution to convert all the bonds (double and triple) to single bond? If not, is-it possible to provide this kind of node?
(Ok, here one solution is to use canonical smiles and to transform this smile code to upper case. But not very usefull).
I was under the impression that carbonyls should be retained in Murcko Scaffolds. If you think about it, is important, as any removal of double bonds (i.e. carbonyl) would affect the shape of the scaffold. Same is true for amide links between aromatics, these are retained too, and any alkenyl side branches have the first carbon atom retained too.
You will see the same effect with the RDkit Murcko scaffold node as well. Removal of these double bonds will misrepresent the original scaffold shape.
I hope this helps.
After a more extensive bibliography I recognize that the definition used in the Indigo node (and the RDkit node) for the Murcko scaffold is correct (or at least the most used). What I understand is that "my" definition is an intermediate simplification of the murcko scaffold before obtaining the framework representation (or cyclic skeleton).
So no problem with the implementation of the scaffold node. But I think that clearly a node to transform the compounds structure into the framework (or cyclic) skeleton could be a good things.