{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,9,28]],"date-time":"2026-09-28T14:57:36Z","timestamp":1790607456221,"version":"4.1.0"},"reference-count":43,"publisher":"American Chemical Society (ACS)","issue":"18","license":[{"start":{"date-parts":[[2026,9,7]],"date-time":"2026-09-07T00:00:00Z","timestamp":1788739200000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2026,9,7]],"date-time":"2026-09-07T00:00:00Z","timestamp":1788739200000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"},{"start":{"date-parts":[[2026,9,7]],"date-time":"2026-09-07T00:00:00Z","timestamp":1788739200000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-045"}],"funder":[{"DOI":"10.13039\/100014103","name":"Key Technology Research and Development Program of Shandong Province","doi-asserted-by":"publisher","award":["2025CXPT121"],"award-info":[{"award-number":["2025CXPT121"]}],"id":[{"id":"10.13039\/100014103","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100010040","name":"Taishan Scholar Project of Shandong Province","doi-asserted-by":"publisher","award":["tstp.20230606"],"award-info":[{"award-number":["tstp.20230606"]}],"id":[{"id":"10.13039\/501100010040","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100018537","name":"National Science and Technology Major Project","doi-asserted-by":"publisher","award":["2023ZD0507700"],"award-info":[{"award-number":["2023ZD0507700"]}],"id":[{"id":"10.13039\/501100018537","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100018537","name":"National Science and Technology Major Project","doi-asserted-by":"publisher","award":["2023ZD0507704"],"award-info":[{"award-number":["2023ZD0507704"]}],"id":[{"id":"10.13039\/501100018537","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/na","name":"National Natural Science Foundation of China","doi-asserted-by":"publisher","award":["22377068"],"award-info":[{"award-number":["22377068"]}],"id":[{"id":"10.13039\/na","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/na","name":"National Natural Science Foundation of China","doi-asserted-by":"publisher","award":["22477070"],"award-info":[{"award-number":["22477070"]}],"id":[{"id":"10.13039\/na","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/na","name":"Independent Project of Shandong Basic Research Special Zone","doi-asserted-by":"publisher","award":["TQ052025001"],"award-info":[{"award-number":["TQ052025001"]}],"id":[{"id":"10.13039\/na","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2026,9,28]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:p>The performance of target-specific, ligand-based virtual screening models is strongly influenced by dataset characteristics, including data availability, class imbalance, and evaluation strategies. In this work, we perform a systematic evaluation of graph neural networks (GNNs) using a ChEMBL-derived dataset spanning 5,368 targets and 1.59 million activity records, capturing the long-tailed distributions and target-specific imbalances commonly observed in pharmaceutical data. Through a systematic evaluation of multiple GNN architectures, we identify guidelines for model selection: while the Graph Isomorphism Network (GIN) consistently outperforms others on datasets with &amp;gt;100 samples (a mean ROC\u2013AUC up to 0.94), simpler architectures are more robust under extreme data scarcity. Critically, our comparative analysis of splitting strategies reveals that random sampling yields artificially optimistic performance due to structural overlaps, whereas similarity-aware clustering exposes a substantial generalization gap (AUC drop &amp;gt; 0.3), cautioning against prevailing evaluation practices. We further demonstrate that multi-task learning serves as an effective remedy for small-target instability, providing significant and consistent performance gains. To underscore its translational value, we deploy this comprehensive framework in a virtual screening campaign against Mcl-1, yielding a chemically optimized lead, C4 (Ki = 0.58 \u03bcM), with verified cellular efficacy. Our findings highlight the importance of task-aware benchmark design and offer a practical strategy for reliable GNN application in drug discovery.<\/jats:p>","DOI":"10.1021\/acs.jcim.6c02160","type":"journal-article","created":{"date-parts":[[2026,9,7]],"date-time":"2026-09-07T20:01:41Z","timestamp":1788811301000},"page":"11756-11766","source":"Crossref","is-referenced-by-count":0,"title":["Systematic Evaluation\nof Graph Neural Networks for\nLigand-Based Virtual Screening on ChEMBL Datasets"],"prefix":"10.1021","volume":"66","author":[{"given":"Haihan","family":"Liu","sequence":"first","affiliation":[{"name":"Shandong University , , ,","place":["Jinan, China, 250012"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jiaqi","family":"Lin","sequence":"additional","affiliation":[{"name":"Shandong University , , ,","place":["Jinan, China, 250012"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ying","family":"Fan","sequence":"additional","affiliation":[{"name":"Shandong University , , ,","place":["Jinan, China, 250012"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jintong","family":"Du","sequence":"additional","affiliation":[{"name":"Shandong University , , ,","place":["Jinan, China, 250012"]},{"name":"Shandong First Medical University and Shandong Academy of Medical Sciences , , , ,","place":["Ji\u2019nan, Shandong, China, 250117"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Xinying","family":"Yang","sequence":"additional","affiliation":[{"name":"Shandong University , , ,","place":["Jinan, China, 250012"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-2879-5146","authenticated-orcid":true,"given":"Hao","family":"Fang","sequence":"additional","affiliation":[{"name":"Shandong University , , ,","place":["Jinan, China, 250012"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-8346-9001","authenticated-orcid":true,"given":"Xuben","family":"Hou","sequence":"additional","affiliation":[{"name":"Shandong University , , ,","place":["Jinan, China, 250012"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"316","published-online":{"date-parts":[[2026,9,7]]},"reference":[{"issue":"14","key":"2026092810075078400_cit1","doi-asserted-by":"publisher","first-page":"5413","DOI":"10.1021\/acs.jcim.4c00072","article-title":"Docking Score ML: Target-Specific\nMachine Learning Models Improving Docking-Based Virtual Screening\nin 155 Targets","volume":"64","author":"Liu","year":"2024","journal-title":"J. Chem. Inf. Model."},{"issue":"4","key":"2026092810075078400_cit2","doi-asserted-by":"publisher","first-page":"108","DOI":"10.1007\/s00894-023-05507-6","article-title":"Structural\nRequirement of RAR\u03b3 Agonism through Computational Aspects","volume":"29","author":"Liu","year":"2023","journal-title":"J. Mol. Model."},{"key":"2026092810075078400_cit3","doi-asserted-by":"publisher","first-page":"107864","DOI":"10.1016\/j.compbiolchem.2023.107864","article-title":"Computational Characteristics of the Structure-Activity\nRelationship\nof Inhibitors Targeting Pks13-TE Domain","volume":"104","author":"Wang","year":"2023","journal-title":"Comput.\nBiol. Chem."},{"issue":"Database","key":"2026092810075078400_cit4","doi-asserted-by":"publisher","first-page":"D198","DOI":"10.1093\/nar\/gkl999","article-title":"BindingDB: A Web-Accessible\nDatabase of Experimentally Determined Protein-Ligand Binding Affinities","volume":"35","author":"Liu","year":"2007","journal-title":"Nucleic Acids Res."},{"issue":"1","key":"2026092810075078400_cit5","doi-asserted-by":"publisher","first-page":"63","DOI":"10.1007\/s42514-021-00086-5","article-title":"Molecular Docking-Based Computational\nPlatform for High-Throughput Virtual Screening","volume":"4","author":"Zhang","year":"2022","journal-title":"CCF Trans. HPC"},{"issue":"2","key":"2026092810075078400_cit6","doi-asserted-by":"publisher","first-page":"127","DOI":"10.1038\/s42256-021-00438-4","article-title":"Geometry-Enhanced Molecular\nRepresentation Learning for Property Prediction","volume":"4","author":"Fang","year":"2022","journal-title":"Nat. Mach. Intell."},{"key":"2026092810075078400_cit7","article-title":"A Survey of Graph Neural Networks\nfor Drug Discovery: Recent Developments and Challenges","author":"Berry","year":"2025","journal-title":"arXiv"},{"issue":"11","key":"2026092810075078400_cit8","doi-asserted-by":"publisher","first-page":"1859","DOI":"10.1002\/jcc.20945","article-title":"CHARMM-GUI: A Web-based Graphical\nUser Interface for CHARMM","volume":"29","author":"Jo","year":"2008","journal-title":"J. Comput. Chem."},{"issue":"1","key":"2026092810075078400_cit9","doi-asserted-by":"publisher","first-page":"20","DOI":"10.1186\/s13321-024-00812-5","article-title":"Reinvent 4: Modern AI\u2013Driven Generative Molecule\nDesign","volume":"16","author":"Loeffler","year":"2024","journal-title":"J. Cheminform."},{"issue":"6","key":"2026092810075078400_cit10","doi-asserted-by":"publisher","first-page":"bbac408","DOI":"10.1093\/bib\/bbac408","article-title":"FP-GNN: A\nVersatile Deep Learning Architecture for Enhanced Molecular Property\nPrediction","volume":"23","author":"Cai","year":"2022","journal-title":"Briefings Bioinf."},{"issue":"4","key":"2026092810075078400_cit11","doi-asserted-by":"publisher","first-page":"2369","DOI":"10.1021\/acs.jcim.6c00014","article-title":"WeMol: A Cloud-Based\nand Zero-Code Platform for AI-Driven Molecular Design and Simulation","volume":"66","author":"Liu","year":"2026","journal-title":"J. Chem. Inf. Model."},{"issue":"14","key":"2026092810075078400_cit12","doi-asserted-by":"publisher","first-page":"6582","DOI":"10.1021\/jm300687e","article-title":"Directory of Useful\nDecoys, Enhanced (DUD-E): Better Ligands and Decoys for Better Benchmarking","volume":"55","author":"Mysinger","year":"2012","journal-title":"J. Med. Chem."},{"issue":"10","key":"2026092810075078400_cit13","doi-asserted-by":"publisher","first-page":"2650","DOI":"10.1021\/ci2001549","article-title":"DEKOIS: Demanding Evaluation Kits\nfor Objective in Silico Screening\u2500A Versatile Tool for Benchmarking\nDocking Programs and Scoring Functions","volume":"51","author":"Vogel","year":"2011","journal-title":"J. Chem.\nInf. Model."},{"issue":"6","key":"2026092810075078400_cit14","doi-asserted-by":"publisher","first-page":"1447","DOI":"10.1021\/ci400115b","article-title":"Evaluation and Optimization\nof Virtual Screening Workflows with DEKOIS 2.0 \u2013 A Public Library\nof Challenging Docking Benchmark Sets","volume":"53","author":"Bauer","year":"2013","journal-title":"J. Chem.\nInf. Model."},{"issue":"1","key":"2026092810075078400_cit15","doi-asserted-by":"publisher","first-page":"21","DOI":"10.1186\/s13321-015-0074-6","article-title":"Applying DEKOIS 2.0 in Structure-Based\nVirtual Screening to Probe the Impact of Preparation Procedures and\nScore Normalization","volume":"7","author":"Ibrahim","year":"2015","journal-title":"J. Cheminform."},{"issue":"9","key":"2026092810075078400_cit16","doi-asserted-by":"publisher","first-page":"4263","DOI":"10.1021\/acs.jcim.0c00155","article-title":"LIT-PCBA: An Unbiased Data Set for\nMachine Learning and Virtual Screening","volume":"60","author":"Tran-Nguyen","year":"2020","journal-title":"J. Chem.\nInf. Model."},{"issue":"8","key":"2026092810075078400_cit17","doi-asserted-by":"publisher","DOI":"10.1371\/journal.pone.0220113","article-title":"Hidden Bias\nin the\nDUD-E Dataset Leads to Misleading Performance of Deep Learning in\nStructure-Based Virtual Screening","volume":"14","author":"Chen","year":"2019","journal-title":"PLoS One"},{"issue":"D1","key":"2026092810075078400_cit18","doi-asserted-by":"publisher","first-page":"D1100","DOI":"10.1093\/nar\/gkr777","article-title":"ChEMBL:\nA Large-Scale Bioactivity Database for Drug Discovery","volume":"40","author":"Gaulton","year":"2012","journal-title":"Nucleic Acids Res."},{"issue":"D1","key":"2026092810075078400_cit19","doi-asserted-by":"publisher","first-page":"D1180","DOI":"10.1093\/nar\/gkad1004","article-title":"The ChEMBL Database in 2023: A Drug\nDiscovery Platform Spanning Multiple Bioactivity Data Types and Time\nPeriods","volume":"52","author":"Zdrazil","year":"2024","journal-title":"Nucleic Acids Res."},{"key":"2026092810075078400_cit20","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-72359-9_5","article-title":"Scaffold Splits Overestimate Virtual Screening Performance","volume-title":"Artificial Neural Networks And Machine Learning \u2013\nICANN 2024","author":"Guo","year":"2024"},{"issue":"1","key":"2026092810075078400_cit21","doi-asserted-by":"publisher","first-page":"96","DOI":"10.1186\/s13321-021-00576-2","article-title":"Splitting\nChemical Structure Data Sets for Federated Privacy-Preserving Machine\nLearning","volume":"13","author":"Simm","year":"2021","journal-title":"J. Cheminform."},{"issue":"1","key":"2026092810075078400_cit22","doi-asserted-by":"publisher","first-page":"94","DOI":"10.1186\/s13321-025-01039-8","article-title":"UMAP-Based Clustering Split for Rigorous\nEvaluation of AI Models for Virtual Screening on Cancer Cell Lines*","volume":"17","author":"Guo","year":"2025","journal-title":"J. Cheminform."},{"issue":"2","key":"2026092810075078400_cit23","doi-asserted-by":"publisher","first-page":"513","DOI":"10.1039\/C7SC02664A","article-title":"MoleculeNet: A Benchmark\nfor Molecular Machine Learning","volume":"9","author":"Wu","year":"2018","journal-title":"Chem. Sci."},{"issue":"4","key":"2026092810075078400_cit24","doi-asserted-by":"publisher","first-page":"783","DOI":"10.1021\/ci400084k","article-title":"Time-Split\nCross-Validation as a Method for Estimating the Goodness of Prospective\nPrediction","volume":"53","author":"Sheridan","year":"2013","journal-title":"J. Chem. Inf. Model."},{"issue":"1","key":"2026092810075078400_cit25","doi-asserted-by":"publisher","first-page":"123","DOI":"10.1186\/s12915-025-02231-w","article-title":"Negative Sampling\nStrategies Impact the Prediction of Scale-Free Biomolecular Network\nInteractions with Machine Learning","volume":"23","author":"Li","year":"2025","journal-title":"BMC Biol."},{"issue":"4","key":"2026092810075078400_cit26","doi-asserted-by":"publisher","DOI":"10.1371\/journal.pone.0061007","article-title":"Comparability of Mixed\nIC50 Data \u2013 A Statistical Analysis","volume":"8","author":"Kalliokoski","year":"2013","journal-title":"PLoS\nOne"},{"issue":"1","key":"2026092810075078400_cit27","doi-asserted-by":"publisher","first-page":"16","DOI":"10.1186\/s13321-016-0127-5","article-title":"Impact\nof Similarity Threshold on\nthe Topology of Molecular Similarity Networks and Clustering Outcomes","volume":"8","author":"Zahora\u0301nszky-Ko\u030bhalmi","year":"2016","journal-title":"J. Cheminform."},{"key":"2026092810075078400_cit28","article-title":"Semi-Supervised\nClassification with Graph Convolutional Networks","author":"Kipf","year":"2016","journal-title":"arXiv"},{"key":"2026092810075078400_cit29","article-title":"Graph\nAttention Networks","author":"Velic\u030ckovic\u0301","year":"2018","journal-title":"arXiv"},{"key":"2026092810075078400_cit30","article-title":"How Powerful Are Graph Neural\nNetworks?","author":"Xu","year":"2018","journal-title":"arXiv"},{"key":"2026092810075078400_cit31","article-title":"Inductive\nRepresentation\nLearning on Large Graphs","author":"Hamilton","year":"2017","journal-title":"arXiv"},{"issue":"7","key":"2026092810075078400_cit32","doi-asserted-by":"publisher","first-page":"4291","DOI":"10.1021\/acs.jctc.1c00302","article-title":"OPLS4: Improving\nForce Field Accuracy on Challenging Regimes of Chemical Space","volume":"17","author":"Lu","year":"2021","journal-title":"J. Chem. Theory Comput."},{"issue":"5","key":"2026092810075078400_cit33","doi-asserted-by":"publisher","first-page":"2054","DOI":"10.1021\/acs.jmedchem.5b01660","article-title":"Discovery of 2-Indole-Acylsulfonamide\nMyeloid Cell Leukemia 1 (Mcl-1) Inhibitors Using Fragment-Based Methods","volume":"59","author":"Pelz","year":"2016","journal-title":"J. Med. Chem."},{"issue":"15","key":"2026092810075078400_cit34","doi-asserted-by":"publisher","first-page":"1114","DOI":"10.1002\/jcc.24660","article-title":"CHARMM-GUI 10 Years\nfor Biomolecular Modeling and Simulation","volume":"38","author":"Jo","year":"2017","journal-title":"J.\nComput. Chem."},{"issue":"12","key":"2026092810075078400_cit35","doi-asserted-by":"publisher","first-page":"1582","DOI":"10.1158\/2159-8290.CD-18-0387","article-title":"AMG 176, a Selective MCL1 Inhibitor, Is Effective in\nHematologic Cancer Models Alone and in Combination with Established\nTherapies","volume":"8","author":"Caenepeel","year":"2018","journal-title":"Cancer Discovery"},{"issue":"22","key":"2026092810075078400_cit36","doi-asserted-by":"publisher","first-page":"13762","DOI":"10.1021\/acs.jmedchem.0c01234","article-title":"Discovery\nof S64315, a Potent and\nSelective Mcl-1 Inhibitor","volume":"63","author":"Szlavik","year":"2020","journal-title":"J. Med. Chem."},{"issue":"7626","key":"2026092810075078400_cit37","doi-asserted-by":"publisher","first-page":"477","DOI":"10.1038\/nature19830","article-title":"The MCL1 Inhibitor S63845 Is Tolerable\nand Effective in Diverse Cancer Models","volume":"538","author":"Kotschy","year":"2016","journal-title":"Nature"},{"issue":"1","key":"2026092810075078400_cit38","doi-asserted-by":"publisher","first-page":"138","DOI":"10.1016\/j.bmc.2016.10.020","article-title":"Improved Binding Affinities\nof Pyrrolidine Derivatives as Mcl-1 Inhibitors by Modifying Amino\nAcid Side Chains","volume":"25","author":"Wan","year":"2017","journal-title":"Bioorg. Med. Chem."},{"issue":"17","key":"2026092810075078400_cit39","doi-asserted-by":"publisher","first-page":"9154","DOI":"10.1021\/acs.jmedchem.9b02047","article-title":"Discovery\nof a Copper-Based Mcl-1 Inhibitor as an Effective Antitumor Agent","volume":"63","author":"Lu","year":"2020","journal-title":"J. Med. Chem"},{"key":"2026092810075078400_cit40","article-title":"DiffDock: Diffusion\nSteps,\nTwists, and Turns for Molecular Docking","author":"Corso","year":"2022","journal-title":"arXiv"},{"key":"2026092810075078400_cit41","doi-asserted-by":"publisher","first-page":"1017","DOI":"10.1101\/2024.06.03.597263","article-title":"ECloudGen:\nLeveraging Electron Clouds as a Latent Variable to Scale\nup Structure-Based Molecular Design","volume":"5","author":"Zhang","year":"2025","journal-title":"Nat. Comput.\nSci."},{"issue":"10","key":"2026092810075078400_cit42","doi-asserted-by":"publisher","first-page":"860","DOI":"10.1038\/s43588-023-00529-9","article-title":"Computing\nthe Relative Binding Affinity of Ligands Based on a Pairwise Binding\nComparison Network","volume":"3","author":"Yu","year":"2023","journal-title":"Nat. Comput. Sci."},{"issue":"8","key":"2026092810075078400_cit43","doi-asserted-by":"publisher","first-page":"1299","DOI":"10.1038\/s41589-026-02241-x","article-title":"Atomic-Level Protein\u2013Ligand\nRecognition with PBCNet2.0 for Probe Discovery","volume":"22","author":"Yu","year":"2026","journal-title":"Nat. Chem. Biol."}],"container-title":["Journal of Chemical Information\nand Modeling"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/pubs.acs.org\/doi\/pdf\/10.1021\/acs.jcim.6c02160","content-type":"application\/pdf","content-version":"vor","intended-application":"unspecified"},{"URL":"https:\/\/pubs.acs.org\/jcisd8\/article-pdf\/66\/18\/11756\/68432096\/acs.jcim.6c02160.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/pubs.acs.org\/jcisd8\/article-pdf\/66\/18\/11756\/68432096\/acs.jcim.6c02160.pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,9,28]],"date-time":"2026-09-28T14:15:52Z","timestamp":1790604952000},"score":1,"resource":{"primary":{"URL":"https:\/\/pubs.acs.org\/jcisd8\/article\/66\/18\/11756\/5425216\/Systematic-Evaluation-of-Graph-Neural-Networks-for"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,9,7]]},"references-count":43,"journal-issue":{"issue":"18","published-online":{"date-parts":[[2026,9,7]]},"published-print":{"date-parts":[[2026,9,28]]}},"URL":"https:\/\/doi.org\/10.1021\/acs.jcim.6c02160","relation":{},"ISSN":["1549-9596","1549-960X"],"issn-type":[{"value":"1549-9596","type":"print"},{"value":"1549-960X","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,9,7]]}}}