Mining significant substructure pairs for interpreting polypharmacology in drug-target network.

1c ): This test was performed because 1) frequent substructure pairs in drug-target pairs may frequently appear in non-interacting pairs and in this case they cannot be significant and 2) if any substructure of a substructure pair is already frequently observed in drug-target pairs rather than in non-interacting pairs, this substructure pair might not be significant even if it appears more frequently in drug-target pairs than non-interacting pairs.

also in 673 other papers