Table 3.

Nearest-neighbor-based overlap between datasets.

Descriptor block	Dataset	%NN in PDB	%NN in DrugBank
Protein	DrugBank	20	80
Protein	PDB	94	6
Ligand	DrugBank	19	81
Ligand	PDB	93	7
Protein-ligand	DrugBank	39	61
Protein-ligand	PDB	86	14

This table contains the percentage nearest-neighbor (NN) in the tree PCA models based on protein descriptors, ligand descriptors and protein-ligand descriptors. The NN overlap is reported for DrugBank vs. PDB and vice versa.