R-group generation

[gry-613]

Hello, when using the LibInvent module for R-group generation, I noticed that the generated structures are mostly large, ring-containing ones. Is there any way to make the generated R-groups smaller?
And I was wondering if the modifying groups in the training data are predominantly large ring structures? If possible, could you also share the rules you used to separate scaffolds from modifying groups?

[halx]

It would be a great feature to have separate scoring funtions for the fragments as we do in the case of Linkinvent (the complication with Libinvent is that there may be up to 4 fragements and each you may want to treat differenttly). Unfortunately, we do not have that yet.

The slicing rules have been published in the original Libinvent paper, which you can find through AUTHORS.md.