We want to measure the selectivity of some docking program recently (supposed to have been) improved. It can reproduce rather well binding modes for most PDB files. Now we would like to try its affinity prediction correlation for a small set of ligands versus different receptors. It is called cross docking or inverse docking or receptorome screening. I wonder which databases containing affinity data for this kind of sets you can recommend

P.S. I have found PDBBind database which is fine since it contains many experimental data. My problem now is how can I get data following the condition "one ligand that binds with several receptors". At the moment I am trying; I go to PDB and as I can search by ligand, try to be lucky and try several ligands. For ligands which participate in several receptors, then I try to find affinity information for it, in the PDBBind database. Do you know a better approach?

P.S.2: In another way, which are the most frequent ligands in the PDB database? Answers please here.