{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,6,14]],"date-time":"2026-06-14T23:53:40Z","timestamp":1781481220730,"version":"3.54.1"},"reference-count":34,"publisher":"Springer Science and Business Media LLC","issue":"1","license":[{"start":{"date-parts":[[2026,4,28]],"date-time":"2026-04-28T00:00:00Z","timestamp":1777334400000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"},{"start":{"date-parts":[[2026,6,15]],"date-time":"2026-06-15T00:00:00Z","timestamp":1781481600000},"content-version":"vor","delay-in-days":48,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"}],"content-domain":{"domain":["link.springer.com"],"crossmark-restriction":false},"short-container-title":["Discov Artif Intell"],"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:sec>\n                    <jats:title>Background<\/jats:title>\n                    <jats:p>\n                      Drug repurposing offers a cost-effective alternative to\n                      <jats:italic>de novo<\/jats:italic>\n                      discovery, yet the relative contributions of model complexity, data volume, and feature modalities to knowledge graph\u2013based repurposing remain poorly quantified under rigorous temporal validation.\n                    <\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Methods<\/jats:title>\n                    <jats:p>\n                      We constructed a pharmacology knowledge graph from ChEMBL\u00a036 comprising 5,348 entities (3,127 drugs, 1,156 proteins, 1,065 indications) and 20,015 edges across 4 relation types. We enforced a strict temporal split (training:\n                      <jats:inline-formula>\n                        <jats:tex-math>$$\\le $$<\/jats:tex-math>\n                      <\/jats:inline-formula>\n                      2022; testing: 2023\u20132025) with biologically verified hard negatives mined from failed assays and clinical trials. We benchmarked five Knowledge Graph Embedding models (TransE, TransR, RotatE, ComplEx, DistMult; 0.78\u20130.81M parameters) and a Standard GNN (3.44M parameters) that incorporates drug chemical structure using a GAT encoder and ESM-2 embeddings evaluated by PR-AUC and Hits@\n                      <jats:italic>k<\/jats:italic>\n                      on drug\u2013protein and drug\u2013indication link prediction. Scaling (0.78M\u20139.75M parameters; 25\u2013100% data) and feature ablation studies isolated contributions of model capacity, graph density, and node feature modalities.\n                    <\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>Feature ablation revealed a counter-intuitive performance hierarchy. Removing the GAT-based drug structure encoder entirely from the GNN and retaining only topological embeddings combined with ESM-2 protein features improved drug\u2013protein PR-AUC from 0.5631 to 0.5785, while simultaneously reducing VRAM usage from 5.30 GB to 353 MB. Additionally, replacing the GAT encoder with Morgan fingerprints further degraded performance (PR-AUC = 0.5286), indicating that explicit chemical structure representations can be not only redundant but detrimental for predicting pharmacological network interactions. Scaling the GNN beyond 2.44 M parameters yielded diminishing performance gains, whereas increasing training data consistently improved model performance with no observable ceiling. External validation confirmed 6 of the top 14 novel predictions (42.9%) as established therapeutic indications.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Conclusions<\/jats:title>\n                    <jats:p>Drug pharmacological behavior can be accurately predicted using target-centric information and drug\u2013network topology alone, without requiring explicit drug chemical structure representations. Model performance is substantially more sensitive to data volume and node density than to architectural complexity, with scaling in model size yielding limited returns relative to improvements in graph coverage. Consequently, state-of-the-art performance is achievable on budget hardware, with a model using only 352 MB VRAM on a consumer GPU.<\/jats:p>\n                  <\/jats:sec>","DOI":"10.1007\/s44163-026-01303-2","type":"journal-article","created":{"date-parts":[[2026,4,28]],"date-time":"2026-04-28T15:29:36Z","timestamp":1777390176000},"update-policy":"https:\/\/doi.org\/10.1007\/springer_crossmark_policy","source":"Crossref","is-referenced-by-count":0,"title":["Pharmacology knowledge graphs enable drug repurposing without chemical structure information"],"prefix":"10.1007","volume":"6","author":[{"given":"Youssef","family":"Abo-Dahab","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ruby","family":"Hernandez","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ismael Caleb Arechiga","family":"Duran","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"297","published-online":{"date-parts":[[2026,4,28]]},"reference":[{"key":"1303_CR1","doi-asserted-by":"publisher","unstructured":"Pinzi L, Bisi N, Rastelli G. How drug repurposing can advance drug discovery: challenges and opportunities. Frontiers in Drug Discovery. 2024;4. Mini Review, Section: In silico Methods and Artificial Intelligence for Drug Discovery. https:\/\/doi.org\/10.3389\/fddsv.2024.1460100.","DOI":"10.3389\/fddsv.2024.1460100"},{"issue":"16","key":"1303_CR2","doi-asserted-by":"publisher","first-page":"1923","DOI":"10.2174\/1568026614666140929124445","volume":"14","author":"E Lionta","year":"2014","unstructured":"Lionta E, Spyrou G, Vassilatis DK, Cournia Z. Structure-based virtual screening for drug discovery: principles, applications and recent advances. Curr Top Med Chem. 2014;14(16):1923\u201338.","journal-title":"Curr Top Med Chem"},{"key":"1303_CR3","doi-asserted-by":"publisher","first-page":"862","DOI":"10.1038\/nature03197","volume":"432","author":"BK Shoichet","year":"2004","unstructured":"Shoichet BK. Virtual screening of chemical libraries. Nature. 2004;432:862\u20135.","journal-title":"Nature"},{"key":"1303_CR4","unstructured":"Abo-Dahab Y, Arzamassky A.: Entresto (Sacubitril\/Valsartan): How one company managed to have a monopoly on the most important drug for heart failure. Zenodo. Available from: https:\/\/doi.org\/10.5281\/zenodo.14911774."},{"issue":"1","key":"1303_CR5","doi-asserted-by":"publisher","first-page":"328","DOI":"10.1186\/s12915-025-02433-2","volume":"23","author":"N Almusallam","year":"2025","unstructured":"Almusallam N, Khan S, Alarfaj FK, Ahmad N. A robust deep learning framework for RNA 5-methyluridine modification prediction using integrated features. BMC Biol. 2025;23(1):328. https:\/\/doi.org\/10.1186\/s12915-025-02433-2.","journal-title":"BMC Biol"},{"issue":"1","key":"1303_CR6","doi-asserted-by":"publisher","first-page":"35872","DOI":"10.1038\/s41598-025-19689-x","volume":"15","author":"S Khan","year":"2025","unstructured":"Khan S, Dilshad N, Ahmad N, Noor S, AlQahtani SA. Integrating AI in security information and event management for real time cyber defense. Sci Rep. 2025;15(1):35872. https:\/\/doi.org\/10.1038\/s41598-025-19689-x.","journal-title":"Sci Rep"},{"issue":"9","key":"1303_CR7","doi-asserted-by":"publisher","first-page":"1057","DOI":"10.1080\/17460441.2021.1910673","volume":"16","author":"F MacLean","year":"2021","unstructured":"MacLean F. Knowledge graphs and their applications in drug discovery. Expert Opin Drug Discov. 2021;16(9):1057\u201369. https:\/\/doi.org\/10.1080\/17460441.2021.1910673.","journal-title":"Expert Opin Drug Discov"},{"key":"1303_CR8","doi-asserted-by":"publisher","unstructured":"Himmelstein DS, Baranzini SE. Systematic integration of biomedical knowledge prioritizes drugs for repurposing. eLife. 2017;6:e26726. https:\/\/doi.org\/10.7554\/eLife.26726.","DOI":"10.7554\/eLife.26726"},{"key":"1303_CR9","unstructured":"Bordes A, Usunier N, Garcia-Duran A, Weston J, Yakhnenko O. Translating Embeddings for Modeling Multi-Relational Data. In: Advances in Neural Information Processing Systems. vol.\u00a026; 2013. ."},{"key":"1303_CR10","doi-asserted-by":"crossref","unstructured":"Lin Y, Liu Z, Sun M, Liu Y, Zhu X. Learning entity and relation embeddings for knowledge graph completion. In Proceedings of the 29th AAAI conference on artificial intelligence. 2015.","DOI":"10.1609\/aaai.v29i1.9491"},{"issue":"1","key":"1303_CR11","doi-asserted-by":"publisher","first-page":"61","DOI":"10.1109\/TNN.2008.2005605","volume":"20","author":"F Scarselli","year":"2009","unstructured":"Scarselli F, Gori M, Tsoi AC, Hagenbuchner M, Monfardini G. The graph neural network model. IEEE Trans Neural Netw. 2009;20(1):61\u201380.","journal-title":"IEEE Trans Neural Netw"},{"key":"1303_CR12","unstructured":"Veli\u010dkovi\u0107 P, Cucurull G, Casanova A, Romero A, Li\u00f2 P, Bengio Y. Graph attention networks. In International conference on learning representations. 2018. Available from: https:\/\/openreview.net\/forum?id=rJXMpikCZ."},{"key":"1303_CR13","unstructured":"Yun S, Jeong M, Kim R, Kang J, Kim HJ. Graph transformer networks. In Advances in neural information processing systems. 2019."},{"key":"1303_CR14","unstructured":"Dwivedi VP, Bresson X. A generalization of transformer networks to graphs. arXiv preprint arXiv:2012.09699. 2020;."},{"key":"1303_CR15","unstructured":": ChEMBL Database Release 36. ChEMBL36 (Accessed October 2025), https:\/\/doi.org\/10.6019\/CHEMBL.database.36.https:\/\/www.ebi.ac.uk\/chembl\/."},{"key":"1303_CR16","doi-asserted-by":"publisher","unstructured":"Wang Y, Ruffinelli D, Broscheit S, Gemulla R. On evaluating embedding models for knowledge base completion. arXiv preprint arXiv:1810.07180. 2018;https:\/\/doi.org\/10.48550\/arXiv.1810.07180.","DOI":"10.48550\/arXiv.1810.07180"},{"key":"1303_CR17","unstructured":"U S Food and Drug Administration.: FDA approves Pradaxa to treat deep vein thrombosis and pulmonary embolism. Accessed: 2026-02-22. https:\/\/www.fda.gov\/news-events\/press-announcements\/fda-approves-pradaxa-treat-deep-vein-thrombosis-and-pulmonary-embolism."},{"issue":"9227","key":"1303_CR18","doi-asserted-by":"publisher","first-page":"359","DOI":"10.1016\/S0140-6736(00)02514-9","volume":"356","author":"L Hansson","year":"2000","unstructured":"Hansson L, Hedner T, Lund-Johansen P, Kjeldsen SE, Lindholm LH, Syvertsen JO, et al. Randomised trial of effects of calcium antagonists compared with diuretics and beta-blockers on cardiovascular morbidity and mortality in hypertension: the Nordic Diltiazem (NORDIL) trial. The Lancet. 2000;356(9227):359\u201365. https:\/\/doi.org\/10.1016\/S0140-6736(00)02514-9.","journal-title":"The Lancet"},{"issue":"3","key":"1303_CR19","doi-asserted-by":"publisher","first-page":"149","DOI":"10.1016\/S1474-4422(02)00071-1","volume":"1","author":"PB Gorelick","year":"2002","unstructured":"Gorelick PB. New horizons for stroke prevention: progress and hope. Lancet Neurol. 2002;1(3):149\u201356. https:\/\/doi.org\/10.1016\/S1474-4422(02)00071-1.","journal-title":"Lancet Neurol"},{"key":"1303_CR20","unstructured":"Drugs com.: Cortisone: Uses, Dosage, Side Effects. Accessed: 2026-02-22. https:\/\/www.drugs.com\/cortisone.html."},{"issue":"13","key":"1303_CR21","doi-asserted-by":"publisher","first-page":"818","DOI":"10.1056\/NEJM198503283121303","volume":"312","author":"ME Weinblatt","year":"1985","unstructured":"Weinblatt ME, Coblyn JS, Fox DA, Fraser PA, Holdsworth DE, Glass DN, et al. Efficacy of low-dose methotrexate in rheumatoid arthritis. N Engl J Med. 1985;312(13):818\u201322. https:\/\/doi.org\/10.1056\/NEJM198503283121303.","journal-title":"N Engl J Med"},{"key":"1303_CR22","doi-asserted-by":"publisher","first-page":"175628481876993","DOI":"10.1177\/1756284818769936","volume":"11","author":"V Savarino","year":"2018","unstructured":"Savarino V, Marabotto E, Zentilin P, Furnari M, Bodini G, De Maria C, et al. Appropriate use of proton pump inhibitors: when, what, and for how long? Ther Adv Gastroenterol. 2018;11:1756284818769936. https:\/\/doi.org\/10.1177\/1756284818769936.","journal-title":"Ther Adv Gastroenterol"},{"issue":"7","key":"1303_CR23","doi-asserted-by":"publisher","first-page":"1366","DOI":"10.1111\/jdv.18914","volume":"37","author":"JI Silverberg","year":"2023","unstructured":"Silverberg JI, Bissonnette R, Kircik L, Murrell DF, Selfridge A, Liu K, et al. Efficacy and safety of etrasimod, a sphingosine 1-phosphate receptor modulator, in adults with moderate-to-severe atopic dermatitis (ADVISE). J Eur Acad Dermatol Venereol. 2023;37(7):1366\u201374. https:\/\/doi.org\/10.1111\/jdv.18914.","journal-title":"J Eur Acad Dermatol Venereol"},{"key":"1303_CR24","doi-asserted-by":"publisher","unstructured":"Cattaneo A, Bonner S, Martynec T, Morrissey E, Luschi C, Barrett IP, et\u00a0al. The Role of Graph Topology in the Performance of Biomedical Knowledge Graph Completion Models. Bioinformatics. 2024;41(10):btaf547. https:\/\/doi.org\/10.1093\/bioinformatics\/btaf547.","DOI":"10.1093\/bioinformatics\/btaf547"},{"issue":"1","key":"1303_CR25","doi-asserted-by":"publisher","first-page":"5","DOI":"10.1186\/s13326-024-00308-z","volume":"15","author":"DN Sosa","year":"2024","unstructured":"Sosa DN, Neculae G, Fauqueur J, Altman RB. Elucidating the semantics-topology trade-off for knowledge inference-based pharmacological discovery. J Biomed Semant. 2024;15(1):5. https:\/\/doi.org\/10.1186\/s13326-024-00308-z.","journal-title":"J Biomed Semant"},{"issue":"1","key":"1303_CR26","doi-asserted-by":"publisher","first-page":"88","DOI":"10.1186\/s12859-025-06101-8","volume":"26","author":"S Khan","year":"2025","unstructured":"Khan S, Noor S, Awan HH, Iqbal S, AlQahtani SA, Dilshad N, et al. Deep-ProBind: binding protein prediction with transformer-based deep learning model. BMC Bioinf. 2025;26(1):88. https:\/\/doi.org\/10.1186\/s12859-025-06101-8.","journal-title":"BMC Bioinf"},{"key":"1303_CR27","doi-asserted-by":"publisher","first-page":"1596216","DOI":"10.3389\/fphar.2025.1596216","volume":"16","author":"X Jiang","year":"2025","unstructured":"Jiang X, Wen L, Li W, Que D, Ming L. DTGHAT: multi-molecule heterogeneous graph transformer based on multi-molecule graph for drug-target identification. Front Pharmacol. 2025;16:1596216. https:\/\/doi.org\/10.3389\/fphar.2025.1596216.","journal-title":"Front Pharmacol"},{"key":"1303_CR28","doi-asserted-by":"crossref","unstructured":"Xiang C, Ma T, Fu X, Liu Y, Song B, Zeng X. From knowledge to treatment: large language model assisted biomedical concept representation for drug repurposing. arXiv preprint arXiv:2510.12181. 2025;.","DOI":"10.18653\/v1\/2025.findings-emnlp.751"},{"issue":"12","key":"1303_CR29","doi-asserted-by":"publisher","first-page":"3601","DOI":"10.1038\/s41591-024-03233-x","volume":"30","author":"K Huang","year":"2024","unstructured":"Huang K, Chandak P, Wang Q, Havaldar S, Vaid A, Leskovec J, et al. A foundation model for clinician-centered drug repurposing. Nat Med. 2024;30(12):3601\u201313. https:\/\/doi.org\/10.1038\/s41591-024-03233-x.","journal-title":"Nat Med"},{"key":"1303_CR30","doi-asserted-by":"publisher","unstructured":"Firoozbakht F, S\u00fcwer S, Elkjaer ML, Handy DE, Maier A, Li J, et\u00a0al. NetMedGPT - A network medicine foundation model for extensive disease mechanism mining and drug repurposing. bioRxiv. 2026. https:\/\/doi.org\/10.64898\/2026.01.04.697552.","DOI":"10.64898\/2026.01.04.697552"},{"key":"1303_CR31","unstructured":"Hamilton W, Ying Z, Leskovec J. Inductive representation learning on large graphs. In Advances in neural information processing systems. 2017."},{"key":"1303_CR32","doi-asserted-by":"publisher","unstructured":"Ying R, Bourgeois D, You J, Zitnik M, Leskovec J. GNNExplainer: Generating explanations for graph neural networks. arXiv preprint arXiv:1903.03894. 2019;https:\/\/doi.org\/10.48550\/arXiv.1903.03894.","DOI":"10.48550\/arXiv.1903.03894"},{"key":"1303_CR33","doi-asserted-by":"publisher","unstructured":"Zeng H, Zhou H, Srivastava A, Kannan R, Prasanna V. GraphSAINT: Graph sampling based inductive learning method. arXiv preprint arXiv:1907.04931. 2019;https:\/\/doi.org\/10.48550\/arXiv.1907.04931.","DOI":"10.48550\/arXiv.1907.04931"},{"key":"1303_CR34","doi-asserted-by":"publisher","unstructured":"Chiang WL, Liu X, Si S, Li Y, Bengio S, Hsieh CJ. Cluster-GCN: An efficient algorithm for training deep and large graph convolutional networks. arXiv preprint arXiv:1905.07953. 2019;https:\/\/doi.org\/10.48550\/arXiv.1905.07953.","DOI":"10.48550\/arXiv.1905.07953"}],"container-title":["Discover Artificial Intelligence"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/link.springer.com\/article\/10.1007\/s44163-026-01303-2","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/link.springer.com\/content\/pdf\/10.1007\/s44163-026-01303-2.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/link.springer.com\/content\/pdf\/10.1007\/s44163-026-01303-2.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,6,14]],"date-time":"2026-06-14T23:48:07Z","timestamp":1781480887000},"score":1,"resource":{"primary":{"URL":"https:\/\/link.springer.com\/10.1007\/s44163-026-01303-2"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,4,28]]},"references-count":34,"journal-issue":{"issue":"1","published-online":{"date-parts":[[2026,12]]}},"alternative-id":["1303"],"URL":"https:\/\/doi.org\/10.1007\/s44163-026-01303-2","relation":{},"ISSN":["2731-0809"],"issn-type":[{"value":"2731-0809","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,4,28]]},"assertion":[{"value":"20 January 2026","order":1,"name":"received","label":"Received","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"15 April 2026","order":2,"name":"accepted","label":"Accepted","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"28 April 2026","order":3,"name":"first_online","label":"First Online","group":{"name":"ArticleHistory","label":"Article History"}},{"order":1,"name":"Ethics","group":{"name":"EthicsHeading","label":"Declarations"}},{"value":"Not applicable. This study is a purely computational analysis of publicly available data from the ChEMBL database and does not involve human participants, animal subjects, or any identifiable personal data. No human participants were involved in this study.","order":2,"name":"Ethics","group":{"name":"EthicsHeading","label":"Ethics approval and consent to participate"}},{"value":"Not applicable.","order":3,"name":"Ethics","group":{"name":"EthicsHeading","label":"Consent for publications"}},{"value":"The authors declare no conflict of interest.","order":4,"name":"Ethics","group":{"name":"EthicsHeading","label":"Competing interests"}}],"article-number":"541"}}