{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,26]],"date-time":"2026-10-26T20:22:51Z","timestamp":1793046171572,"version":"4.5.0"},"reference-count":41,"publisher":"Oxford University Press (OUP)","issue":"15","license":[{"start":{"date-parts":[[2021,2,3]],"date-time":"2021-02-03T00:00:00Z","timestamp":1612310400000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"DOI":"10.13039\/501100000266","name":"Engineering and Physical Sciences Research Council","doi-asserted-by":"publisher","award":["EP\/N509711\/1"],"award-info":[{"award-number":["EP\/N509711\/1"]}],"id":[{"id":"10.13039\/501100000266","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2021,8,9]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:sec>\n                    <jats:title>Motivation<\/jats:title>\n                    <jats:p>An essential step in the development of virtual screening methods is the use of established sets of actives and decoys for benchmarking and training. However, the decoy molecules in commonly used sets are biased meaning that methods often exploit these biases to separate actives and decoys, and do not necessarily learn to perform molecular recognition. This fundamental issue prevents generalization and hinders virtual screening method development.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>We have developed a deep learning method (DeepCoy) that generates decoys to a user\u2019s preferred specification in order to remove such biases or construct sets with a defined bias. We validated DeepCoy using two established benchmarks, DUD-E and DEKOIS 2.0. For all 102 DUD-E targets and 80 of the 81 DEKOIS 2.0 targets, our generated decoy molecules more closely matched the active molecules\u2019 physicochemical properties while introducing no discernible additional risk of false negatives. The DeepCoy decoys improved the Deviation from Optimal Embedding (DOE) score by an average of 81% and 66%, respectively, decreasing from 0.166 to 0.032 for DUD-E and from 0.109 to 0.038 for DEKOIS 2.0. Further, the generated decoys are harder to distinguish than the original decoy molecules via docking with Autodock Vina, with virtual screening performance falling from an AUC ROC of 0.70 to 0.63.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Availability and implementation<\/jats:title>\n                    <jats:p>The code is available at https:\/\/github.com\/oxpig\/DeepCoy. Generated molecules can be downloaded from http:\/\/opig.stats.ox.ac.uk\/resources.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Supplementary information<\/jats:title>\n                    <jats:p>Supplementary data are available at Bioinformatics online.<\/jats:p>\n                  <\/jats:sec>","DOI":"10.1093\/bioinformatics\/btab080","type":"journal-article","created":{"date-parts":[[2021,1,28]],"date-time":"2021-01-28T16:30:48Z","timestamp":1611851448000},"page":"2134-2141","source":"Crossref","is-referenced-by-count":101,"title":["Generating property-matched decoy molecules using deep learning"],"prefix":"10.1093","volume":"37","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-6241-0123","authenticated-orcid":false,"given":"Fergus","family":"Imrie","sequence":"first","affiliation":[{"name":"Oxford Protein Informatics Group, Department of Statistics, University of Oxford , Oxford OX1 3LB, UK"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Anthony R","family":"Bradley","sequence":"additional","affiliation":[{"name":"Exscientia Ltd , The Schr\u00f6\u0111inger Building, Oxford Science Park, Oxford OX4 4GE, UK"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-1388-2252","authenticated-orcid":false,"given":"Charlotte M","family":"Deane","sequence":"additional","affiliation":[{"name":"Oxford Protein Informatics Group, Department of Statistics, University of Oxford , Oxford OX1 3LB, UK"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"286","published-online":{"date-parts":[[2021,2,3]]},"reference":[{"key":"2023061310412188300_btab080-B1","doi-asserted-by":"crossref","first-page":"18477","DOI":"10.1073\/pnas.2000585117","article-title":"Machine learning classification can reduce false positives in structure-based virtual screening","volume":"117","author":"Adeshina","year":"2020","journal-title":"Proc. Natl. Acad. Sci, USA"},{"key":"2023061310412188300_btab080-B2","doi-asserted-by":"crossref","first-page":"1447","DOI":"10.1021\/ci400115b","article-title":"Evaluation and optimization of virtual screening workflows with DEKOIS 2.0 \u2013 a public library of challenging docking benchmark sets","volume":"53","author":"Bauer","year":"2013","journal-title":"J. Chem. Inf. Model"},{"key":"2023061310412188300_btab080-B3","doi-asserted-by":"crossref","first-page":"1083","DOI":"10.1093\/nar\/gkt1031","article-title":"The ChEMBL bioactivity database: an update","volume":"42","author":"Bento","year":"2014","journal-title":"Nucleic Acids Res"},{"key":"2023061310412188300_btab080-B4","doi-asserted-by":"crossref","first-page":"56","DOI":"10.1186\/s13321-016-0167-x","article-title":"Benchmark of four popular virtual screening programs: construction of the active\/decoy dataset remains a major determinant of measured performance","volume":"8","author":"Chaput","year":"2016","journal-title":"J. Cheminf"},{"key":"2023061310412188300_btab080-B5","doi-asserted-by":"crossref","first-page":"e0220113","DOI":"10.1371\/journal.pone.0220113","article-title":"Hidden bias in the DUD-E dataset leads to misleading performance of deep learning in structure-based virtual screening","volume":"14","author":"Chen","year":"2019","journal-title":"PLoS One"},{"key":"2023061310412188300_btab080-B6","doi-asserted-by":"crossref","first-page":"8","DOI":"10.1186\/1758-2946-1-8","article-title":"Estimation of synthetic accessibility score of drug-like molecules based on molecular complexity and fragment contributions","volume":"1","author":"Ertl","year":"2009","journal-title":"J. Cheminf"},{"key":"2023061310412188300_btab080-B7","doi-asserted-by":"crossref","first-page":"268","DOI":"10.1021\/acscentsci.7b00572","article-title":"Automatic chemical design using a data-driven continuous representation of molecules","volume":"4","author":"G\u00f3mez-Bombarelli","year":"2018","journal-title":"ACS Cent. Sci"},{"key":"2023061310412188300_btab080-B8","doi-asserted-by":"crossref","first-page":"6789","DOI":"10.1021\/jm0608356","article-title":"Benchmarking sets for molecular docking","volume":"49","author":"Huang","year":"2006","journal-title":"J. Med. Chem"},{"key":"2023061310412188300_btab080-B9","doi-asserted-by":"crossref","first-page":"2319","DOI":"10.1021\/acs.jcim.8b00350","article-title":"Protein family-specific models using deep neural networks and transfer learning improve virtual screening and highlight the need for more data","volume":"58","author":"Imrie","year":"2018","journal-title":"J. Chem. Inf. Model"},{"key":"2023061310412188300_btab080-B10","doi-asserted-by":"crossref","first-page":"1983","DOI":"10.1021\/acs.jcim.9b01120","article-title":"Deep generative models for 3D linker design","volume":"60","author":"Imrie","year":"2020","journal-title":"J. Chem. Inf. Model"},{"key":"2023061310412188300_btab080-B11","author":"Jin","year":"2019"},{"key":"2023061310412188300_btab080-B12","doi-asserted-by":"crossref","first-page":"1893","DOI":"10.1021\/ci300604z","article-title":"Lessons learned in empirical scoring with smina from the CSAR 2011 benchmarking exercise","volume":"53","author":"Koes","year":"2013","journal-title":"J. Chem. Inf. Model"},{"key":"2023061310412188300_btab080-B13","doi-asserted-by":"crossref","first-page":"1297","DOI":"10.1021\/acs.jcim.5b00090","article-title":"Benchmarking data sets for the evaluation of virtual ligand screening methods: review and perspectives","volume":"55","author":"Lagarde","year":"2015","journal-title":"J. Chem. Inf. Model"},{"key":"2023061310412188300_btab080-B14","author":"Landrum","year":"2006"},{"key":"2023061310412188300_btab080-B15","doi-asserted-by":"crossref","first-page":"e1478","DOI":"10.1002\/wcms.1478","article-title":"Machine-learning scoring functions for structure-based virtual screening","volume":"11","author":"Li","year":"2020","journal-title":"WIREs Comput. Mol. Sci"},{"key":"2023061310412188300_btab080-B16","volume-title":"In:","author":"Li","year":"2016"},{"key":"2023061310412188300_btab080-B17","first-page":"7795","volume-title":"Advances in Neural Information Processing Systems 31 (NeurIPS)","author":"Liu","year":"2018"},{"key":"2023061310412188300_btab080-B18","doi-asserted-by":"crossref","first-page":"282","DOI":"10.1021\/acs.jcim.8b00363","article-title":"Practical model selection for prospective virtual screening","volume":"59","author":"Liu","year":"2019","journal-title":"J. Chem. Inf. Model"},{"key":"2023061310412188300_btab080-B19","doi-asserted-by":"crossref","first-page":"224","DOI":"10.1038\/s41586-019-0917-9","article-title":"Ultra-large library docking for discovering new chemotypes","volume":"566","author":"Lyu","year":"2019","journal-title":"Nature"},{"key":"2023061310412188300_btab080-B20","doi-asserted-by":"crossref","first-page":"6582","DOI":"10.1021\/jm300687e","article-title":"Directory of useful decoys, enhanced (DUD-E): Better ligands and decoys for better benchmarking","volume":"55","author":"Mysinger","year":"2012","journal-title":"J. Med. Chem"},{"key":"2023061310412188300_btab080-B21","doi-asserted-by":"crossref","first-page":"239","DOI":"10.1007\/s10822-008-9170-2","article-title":"What do we know and when do we know it?","volume":"22","author":"Nicholls","year":"2008","journal-title":"J. Comput. Aided Mol. Des"},{"key":"2023061310412188300_btab080-B22","doi-asserted-by":"crossref","first-page":"675","DOI":"10.1007\/s10822-013-9672-4","article-title":"Estimation of the size of drug-like chemical space based on GDB-17 data","volume":"27","author":"Polishchuk","year":"2013","journal-title":"J. Comput. Aided Mol. Des"},{"key":"2023061310412188300_btab080-B23","doi-asserted-by":"crossref","first-page":"942","DOI":"10.1021\/acs.jcim.6b00740","article-title":"Protein-ligand scoring with convolutional neural networks","volume":"57","author":"Ragoza","year":"2017","journal-title":"J. Chem. Inf. Model"},{"key":"2023061310412188300_btab080-B24","doi-asserted-by":"crossref","first-page":"11","DOI":"10.3389\/fphar.2018.00011","article-title":"Decoys selection in benchmarking datasets: overview and perspectives","volume":"9","author":"R\u00e9au","year":"2018","journal-title":"Front. Pharmacol"},{"key":"2023061310412188300_btab080-B25","doi-asserted-by":"crossref","first-page":"26","DOI":"10.1186\/1758-2946-5-26","article-title":"Open-source platform to benchmark fingerprints for ligand-based virtual screening","volume":"5","author":"Riniker","year":"2013","journal-title":"J. Cheminf"},{"key":"2023061310412188300_btab080-B26","doi-asserted-by":"crossref","first-page":"742","DOI":"10.1021\/ci100050t","article-title":"Extended-connectivity fingerprints","volume":"50","author":"Rogers","year":"2010","journal-title":"J. Chem. Inf. Model"},{"key":"2023061310412188300_btab080-B27","doi-asserted-by":"crossref","first-page":"169","DOI":"10.1021\/ci8002649","article-title":"Maximum unbiased validation (MUV) data sets for virtual screening based on PubChem bioactivity data","volume":"49","author":"Rohrer","year":"2009","journal-title":"J. Chem. Inf. Model"},{"key":"2023061310412188300_btab080-B28","doi-asserted-by":"crossref","first-page":"120","DOI":"10.1021\/acscentsci.7b00512","article-title":"Generating focused molecule libraries for drug discovery with recurrent neural networks","volume":"4","author":"Segler","year":"2018","journal-title":"ACS Cent. Sci"},{"key":"2023061310412188300_btab080-B29","doi-asserted-by":"crossref","first-page":"947","DOI":"10.1021\/acs.jcim.8b00712","article-title":"In need of bias control: evaluating chemical data for machine learning in structure-based virtual screening","volume":"59","author":"Sieg","year":"2019","journal-title":"J. Chem. Inf. Model"},{"key":"2023061310412188300_btab080-B30","doi-asserted-by":"crossref","first-page":"8867","DOI":"10.1021\/acs.jmedchem.0c00473","article-title":"Discovery of novel inhibitors of a critical brain enzyme using a homology model and a deep convolutional neural network","volume":"63","author":"Stecula","year":"2020","journal-title":"J. Med. Chem"},{"key":"2023061310412188300_btab080-B31","doi-asserted-by":"crossref","first-page":"2324","DOI":"10.1021\/acs.jcim.5b00559","article-title":"ZINC 15 - ligand discovery for everyone","volume":"55","author":"Sterling","year":"2015","journal-title":"J. Chem. Inf. Model"},{"key":"2023061310412188300_btab080-B32","doi-asserted-by":"crossref","first-page":"24817","DOI":"10.1038\/srep24817","article-title":"Constructing and validating high-performance MIEC-SVM models in virtual screening for kinases: a better way for actives discovery","volume":"6","author":"Sun","year":"2016","journal-title":"Sci. Rep"},{"key":"2023061310412188300_btab080-B33","doi-asserted-by":"crossref","first-page":"4263","DOI":"10.1021\/acs.jcim.0c00155","article-title":"LIT-PCBA: an unbiased data set for machine learning and virtual screening","volume":"60","author":"Tran-Nguyen","year":"2020","journal-title":"J. Chem. Inf. Model"},{"key":"2023061310412188300_btab080-B34","doi-asserted-by":"crossref","first-page":"455","DOI":"10.1002\/jcc.21334","article-title":"AutoDock Vina: improving the speed and accuracy of docking with a new scoring function, efficient optimization and multithreading","volume":"31","author":"Trott","year":"2010","journal-title":"J. Comput. Chem"},{"key":"2023061310412188300_btab080-B35","doi-asserted-by":"crossref","first-page":"793","DOI":"10.1021\/ci034289q","article-title":"Virtual screening using protein\u2013ligand docking: avoiding artificial enrichment","volume":"44","author":"Verdonk","year":"2004","journal-title":"J. Chem. Inf. Comput. Sci"},{"key":"2023061310412188300_btab080-B36","doi-asserted-by":"crossref","first-page":"2650","DOI":"10.1021\/ci2001549","article-title":"DEKOIS: demanding evaluation kits for objective in silico screening \u2013 a versatile tool for benchmarking docking programs and scoring functions","volume":"51","author":"Vogel","year":"2011","journal-title":"J. Chem. Inf. Model"},{"key":"2023061310412188300_btab080-B37","doi-asserted-by":"crossref","first-page":"916","DOI":"10.1021\/acs.jcim.7b00403","article-title":"Most ligand-based classification benchmarks reward memorization rather than generalization","volume":"58","author":"Wallach","year":"2018","journal-title":"J. Chem. Inf. Model"},{"key":"2023061310412188300_btab080-B38","doi-asserted-by":"crossref","first-page":"196","DOI":"10.1021\/ci100374f","article-title":"Virtual decoy sets for molecular docking benchmarks","volume":"51","author":"Wallach","year":"2011","journal-title":"J. Chem. Inf. Model"},{"key":"2023061310412188300_btab080-B39","doi-asserted-by":"crossref","first-page":"46710","DOI":"10.1038\/srep46710","article-title":"Performance of machine-learning scoring functions in structure-based virtual screening","volume":"7","author":"W\u00f3jcikowski","year":"2017","journal-title":"Sci. Rep"},{"key":"2023061310412188300_btab080-B40","doi-asserted-by":"crossref","first-page":"5","DOI":"10.4155\/fmc.13.186","article-title":"Challenges and advances in structure-based virtual screening","volume":"6","author":"Yuriev","year":"2014","journal-title":"Fut. Med. Chem"},{"key":"2023061310412188300_btab080-B41","doi-asserted-by":"crossref","first-page":"1038","DOI":"10.1038\/s41587-019-0224-x","article-title":"Deep learning enables rapid identification of potent DDR1 kinase inhibitors","volume":"37","author":"Zhavoronkov","year":"2019","journal-title":"Nat. Biotechnol"}],"container-title":["Bioinformatics"],"original-title":[],"language":"en","link":[{"URL":"http:\/\/academic.oup.com\/bioinformatics\/advance-article-pdf\/doi\/10.1093\/bioinformatics\/btab080\/36297301\/btab080.pdf","content-type":"application\/pdf","content-version":"am","intended-application":"syndication"},{"URL":"https:\/\/academic.oup.com\/bioinformatics\/article-pdf\/37\/15\/2134\/50578579\/btab080.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/academic.oup.com\/bioinformatics\/article-pdf\/37\/15\/2134\/50578579\/btab080.pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2023,10,19]],"date-time":"2023-10-19T09:34:08Z","timestamp":1697708048000},"score":1,"resource":{"primary":{"URL":"https:\/\/academic.oup.com\/bioinformatics\/article\/37\/15\/2134\/6126797"}},"subtitle":[],"editor":[{"given":"Alfonso","family":"Valencia","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"editor"}]}],"short-title":[],"issued":{"date-parts":[[2021,2,3]]},"references-count":41,"journal-issue":{"issue":"15","published-print":{"date-parts":[[2021,8,9]]}},"URL":"https:\/\/doi.org\/10.1093\/bioinformatics\/btab080","relation":{"has-preprint":[{"id-type":"doi","id":"10.1101\/2020.08.26.268193","asserted-by":"object"}]},"ISSN":["1367-4803","1367-4811"],"issn-type":[{"value":"1367-4803","type":"print"},{"value":"1367-4811","type":"electronic"}],"subject":[],"published-other":{"date-parts":[[2021,8,1]]},"published":{"date-parts":[[2021,2,3]]}}}