The article presents the DUD (Directory of Useful Decoys) benchmarking set for molecular docking. DUD contains 2950 annotated ligands for 40 different targets, each with 36 decoy molecules that are physically similar but topologically distinct, resulting in a database of 98,266 compounds. The decoys are designed to be chemically distinct from the ligands to avoid bias in docking enrichment calculations. DUD is freely available online at http://blaster.docking.org/dud/. The study evaluates docking performance using enrichment factors, which measure the ability of a docking method to identify true positives among a database of decoys. The results show that DUD provides a more stringent test for evaluating virtual screening performance compared to other databases like the MDDR. DUD decoys are topologically dissimilar to the annotated ligands, making them likely true negatives. The study also highlights the importance of using unbiased decoys in docking screens to avoid artificial enrichment. The DUD database was created by selecting 2,950 ligands from the literature and generating 36 decoy molecules for each ligand from the ZINC database. The decoys were chosen to have similar physical properties but different topological structures. The database was then prepared for docking using automated procedures, including binding site preparation, scoring grid calculation, and docking calculations. The study compares the performance of DUD with other databases, including the MDDR and smaller decoy sets. The results show that DUD provides better enrichment factors than these databases, indicating that the MDDR database may have biases. The study also demonstrates the importance of cross-docking experiments in evaluating docking specificity. The results show that docking performance varies across different targets, with some targets showing poor enrichment. The study highlights the importance of using a diverse set of targets in benchmarking docking methods. The DUD database is considered a valuable resource for evaluating docking performance, as it provides a benchmarking set that is publicly available and unbiased. The study concludes that DUD is a useful benchmarking set for molecular docking and virtual screening.The article presents the DUD (Directory of Useful Decoys) benchmarking set for molecular docking. DUD contains 2950 annotated ligands for 40 different targets, each with 36 decoy molecules that are physically similar but topologically distinct, resulting in a database of 98,266 compounds. The decoys are designed to be chemically distinct from the ligands to avoid bias in docking enrichment calculations. DUD is freely available online at http://blaster.docking.org/dud/. The study evaluates docking performance using enrichment factors, which measure the ability of a docking method to identify true positives among a database of decoys. The results show that DUD provides a more stringent test for evaluating virtual screening performance compared to other databases like the MDDR. DUD decoys are topologically dissimilar to the annotated ligands, making them likely true negatives. The study also highlights the importance of using unbiased decoys in docking screens to avoid artificial enrichment. The DUD database was created by selecting 2,950 ligands from the literature and generating 36 decoy molecules for each ligand from the ZINC database. The decoys were chosen to have similar physical properties but different topological structures. The database was then prepared for docking using automated procedures, including binding site preparation, scoring grid calculation, and docking calculations. The study compares the performance of DUD with other databases, including the MDDR and smaller decoy sets. The results show that DUD provides better enrichment factors than these databases, indicating that the MDDR database may have biases. The study also demonstrates the importance of cross-docking experiments in evaluating docking specificity. The results show that docking performance varies across different targets, with some targets showing poor enrichment. The study highlights the importance of using a diverse set of targets in benchmarking docking methods. The DUD database is considered a valuable resource for evaluating docking performance, as it provides a benchmarking set that is publicly available and unbiased. The study concludes that DUD is a useful benchmarking set for molecular docking and virtual screening.