DB name	Year	Download address	Origin of the ligands	Origin of the decoys	No. of targets / No. of classes	Decoy compounds selection	Remarks
Rognan's decoy set (Bissantz et al., 2000)	2000	http://bioinfo-pharma.u-strasbg.fr/labwebsite/download.html	Literature	ACD	2/2	Random selection	Design of decoy sets to evaluate the performance of 3 docking programs and 7 scoring functions
Shoichet's decoy set (McGovern and Shoichet, 2003)	2003	–	MDDR	MDDR	9/4	Remove compounds with unwanted functional groups	Compare VS performance depending on the binding site definition (apo, holo or homology modeled structures)
Li's decoy set (Diller and Li, 2003)	2003	–	Literature	MDDR	6/1	Fit polarity and MW to known kinases inhibitors	Compare decoys and ligands physicochemical properties to select decoys
Jain's decoy set (Jain and Nicholls, 2008)	2006	http://www.jainlab.org/downloads.html	PDBbind	ZINC “drug-like” and Rognan's decoy set	34/7	1,000 random molecules from the ZINC that comply to MW ≤ 500, logP ≤ 5, HBA ≤ 10, HBD ≤ 5 and RB ≤ 12 and Rognan's decoys with RB ≤ 15	Use of 5 physicochemical properties to match decoy sets to ligands sets
Directory of Useful Decoys (DUD) (Huang et al., 2006)	2006	http://dud.docking.org	Literature and PDBbind	ZINC “drug-like”	40/6	Decoys must be Lipinski-compliant. The selection is based on both the topologically dissimilarity to ligands and the fit of physicochemical properties	Largest decoy data set so far (40 proteins) and first attempt to select decoys topologically dissimilar decoys
DUD Clusters (Meyer, 2007)	2008	http://dud.docking.org/clusters/	DUD	–	40/6	–	DUD clusters more relevant for scaffold hopping
WOMBAT Datasets (Meyer, 2007)	2007	http://dud.docking.org/wombat/	WOMBAT	–	13/4	–	Design to decrease the analog bias on 13 of the 40 DUD targets, enrich DUD active data sets with compounds from WOMBAT database
Maximum Unbiased Validation (MUV) (Rohrer and Baumann, 2009)	2009	–	PubChem	PubChem	18/7	Two functions measure the active-active and decoy-active distances using 2D chemical descriptors. Actives with the maximum spread within the active set were chosen and decoys with similar spatial distribution were selected	Ligands and decoys are from biologically actives and inactive compounds, i.e., are true actives and inactives, respectively
DUD LIB	2009	http://dud.docking.org/jahn/	DUD-cluster	DUD	13/4	Subset of the DUD database, with more stringent criteria on MW (≤450) and AlogP (≤4,5), and a minimal number of chemotypes	Initially designed for “scaffold-hopping” studies
Charge Matched DUD	2010	http://dud.docking.org/charge-matched/	DUD	ZINC	40/6	Apply a net charge property match on DUD datasets
REPROVIS-DB	2011	–	Literature	Literature	–	Extracted from previous successful studies	Designed for LBVS only
Virtual Decoy sets (VDS) (Wallach and Lilien, 2011)	2011	http://compbio.cs.toronto.edu/VDS	DUD	ZINC	40/6	Same as DUD, but does not consider synthetic feasibility	Purely virtual decoys, availability is not considered
DEKOIS (Vogel et al., 2011)	2011	http://dekois.com/dekois_orig.html	BindingDB	ZINC	40/6	Class decoys and ligands into “cells” based on 6 physicochemical properties and select the closest decoys based on (1) a weighted physicochemical similarity and (2) a LADS score based on functional fingerprints similarity elaborated from the active set	Original treatment of the physicochemical similarity, and introduce the concept of Latent Active in Decoy set, i.e., false false positives
GPCR Ligand (GLL)/Decoys Database (GDD) (Xia et al., 2014)	2012	http://cavasotto-lab.net/Databases/GDD/	GLIDA and PDB structures and Vilar et al., 2010	ZINC	147/1	Physico-chemical properties fit and topological dissimilarity filter. Final selection based on MW	First extensive database targeting a specific protein family
Decoy Finder (Cereto-Massagué et al., 2012)	2012	http://urvnutrigenomica-ctns.github.io/DecoyFinder/	User	User	–	Same as DUD	Graphical tool to generate decoy data sets with adaptable thresholds for physicochemical properties
DUD Enhanced (DUD-E) (Mysinger et al., 2012)	2012	http://dud.docking.org/r2/	CHEMBL	ZINC	102/8	Physico-chemical properties fit along with a topological dissimilarity filter. Random selection of decoys is then applied	Largest database so far (1,420,433 decoys and 66,695 actives)
DEKOIS 2.0 (Ibrahim et al., 2015a)	2013	http://www.dekois.com	BindingDB	ZINC	81/11	Same as DEKOIS with 3 additional physicochemical properties (nFC, nPC, Ar), a PAINS filter and an improved, weighted LADS score
NRLiSt BDB (Lagarde et al., 2014a)	2014	http://nrlist.drugdesign.fr	CHEMBL	ZINC and DUD-E decoys generator	27/1	Use the DUD-E decoy generation tool	Ligands can be either agonists or antagonists (other actives are removed), depending on the purpose of the study
MUBD-HDACs (Xia et al., 2015)	2015	–	CHEMBL and literature	ZINC	14/1	Select decoys based a weighted physicochemical similarity (6 physicochemical properties are considered), and ensure a random spatial distribution of the decoys (i.e., decoys should be as distant to the other actives as a reference ligand)	Applicable both to SBVS and LBVS strategies, uses ligands with proved bioactivity