{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,8]],"date-time":"2026-10-08T19:55:52Z","timestamp":1791489352100,"version":"4.3.4"},"reference-count":41,"publisher":"Springer Science and Business Media LLC","issue":"1","license":[{"start":{"date-parts":[[2026,6,3]],"date-time":"2026-06-03T00:00:00Z","timestamp":1780444800000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"},{"start":{"date-parts":[[2026,6,3]],"date-time":"2026-06-03T00:00:00Z","timestamp":1780444800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"}],"funder":[{"DOI":"10.13039\/501100002241","name":"Japan Science and Technology Agency","doi-asserted-by":"publisher","award":["JPMJSP2140"],"award-info":[{"award-number":["JPMJSP2140"]}],"id":[{"id":"10.13039\/501100002241","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100008131","name":"Rheinische Friedrich-Wilhelms-Universit\u00e4t Bonn","doi-asserted-by":"crossref","id":[{"id":"10.13039\/501100008131","id-type":"DOI","asserted-by":"crossref"}]}],"content-domain":{"domain":["link.springer.com"],"crossmark-restriction":false},"short-container-title":["J Cheminform"],"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:p>The definition of activity cliffs (ACs) depends on compound similarity and activity difference criteria and on activity data types. ACs are usually defined as pairs or groups of structurally similar compounds or structural analogues that are active against the same target, but have large differences in potency (requiring numerical potency values, preferably equilibrium constants). In addition, ACs have also been defined as pairs of structural analogues that are active or inactive in screening assays. In medicinal chemistry, ACs are of particular interest because they often reveal structure\u2013activity relationship (SAR) determinants during compound optimization. In cheminformatics, ACs present challenging test cases for machine learning (ML) and activity predictions because they represent an extreme form of SAR discontinuity in compound data sets. Given their composition, ACs are notoriously difficult to predict based on chemical structure representations. Various attempts have been made to predict compound pairs forming ACs or the activity of AC compounds via ML, often reporting high accuracy. However, in the absence of data leakage from training to test sets, AC prediction accuracy based on chemical structure is low or moderate at best. As an alternative to structural representations, biological\/functional compound descriptors might be considered such as biological assay profiles, which have been investigated for other compound activity prediction. In this work, we report the prediction of AC compounds based on bioactivity profiles derived from a compound profiling matrix using data partitioning schemes designed to control information or data leakage during ML. Under these stringent conditions, AC compound predictions based on bioactivity profiles often failed. However, we also observed subsets of highly accurate predictions and explored in detail why these predictions succeeded, but others failed. The analysis revealed a critically important role of assay similarity for successful AC compound predictions. Most profile assays did not measurably influence the predictions. By contrast, accurate predictions mostly depended on the presence of one or at most a few profile assays that were similar to test assays. In most cases, these profile assays could be identified and exploited for predictions by nearest neighbor searching, thus putting ML performance into perspective.<\/jats:p>\n                  <jats:p>\n                    <jats:bold>Scientific contribution<\/jats:bold>\n                  <\/jats:p>\n                  <jats:p>As an alternative to the generally difficult prediction of ACs based on chemical structure, we introduce prediction of AC compounds based on bioactivity profiles. We show that accurate activity predictions of AC compounds do not depend on the global information content of assay profiles, but are largely determined by nearest neighbor relationships between profile and test assays.<\/jats:p>","DOI":"10.1186\/s13321-026-01210-9","type":"journal-article","created":{"date-parts":[[2026,6,3]],"date-time":"2026-06-03T05:35:09Z","timestamp":1780464909000},"update-policy":"https:\/\/doi.org\/10.1007\/springer_crossmark_policy","source":"Crossref","is-referenced-by-count":2,"title":["Accurate prediction of activity cliff compounds based on bioactivity profiles depends on assay nearest neighbor relationships"],"prefix":"10.1186","volume":"18","author":[{"given":"Ryuto","family":"Abe","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Tomoyuki","family":"Miyao","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"J\u00fcrgen","family":"Bajorath","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"297","published-online":{"date-parts":[[2026,6,3]]},"reference":[{"key":"1210_CR1","doi-asserted-by":"publisher","first-page":"5571","DOI":"10.1021\/jm0705713","volume":"50","author":"L Peltason","year":"2007","unstructured":"Peltason L, Bajorath J (2007) SAR index: quantifying the nature of structure\u2212activity relationships. J Med Chem 50:5571\u20135578. https:\/\/doi.org\/10.1021\/jm0705713","journal-title":"J Med Chem"},{"key":"1210_CR2","doi-asserted-by":"publisher","first-page":"4189","DOI":"10.1021\/acs.jmedchem.5b00948","volume":"59","author":"AT Maynard","year":"2016","unstructured":"Maynard AT, Roberts CD (2016) Quantifying, visualizing, and monitoring lead optimization. J Med Chem 59:4189\u20134202. https:\/\/doi.org\/10.1021\/acs.jmedchem.5b00948","journal-title":"J Med Chem"},{"key":"1210_CR3","doi-asserted-by":"publisher","first-page":"2932","DOI":"10.1021\/jm201706b","volume":"55","author":"D Stumpfe","year":"2012","unstructured":"Stumpfe D, Bajorath J (2012) Exploring activity cliffs in medicinal chemistry. J Med Chem 55:2932\u20132942. https:\/\/doi.org\/10.1021\/jm201706b","journal-title":"J Med Chem"},{"key":"1210_CR4","doi-asserted-by":"publisher","first-page":"18","DOI":"10.1021\/jm401120g","volume":"57","author":"D Stumpfe","year":"2014","unstructured":"Stumpfe D, Hu Y, Dimova D, Bajorath J (2014) Recent progress in understanding activity cliffs and their utility in medicinal chemistry. J Med Chem 57:18\u201328. https:\/\/doi.org\/10.1021\/jm401120g","journal-title":"J Med Chem"},{"key":"1210_CR5","doi-asserted-by":"publisher","first-page":"14360","DOI":"10.1021\/acsomega.9b02221","volume":"4","author":"D Stumpfe","year":"2019","unstructured":"Stumpfe D, Hu H, Bajorath J (2019) Evolving concept of activity cliffs. ACS Omega 4:14360\u201314368. https:\/\/doi.org\/10.1021\/acsomega.9b02221","journal-title":"ACS Omega"},{"key":"1210_CR6","doi-asserted-by":"publisher","first-page":"2181","DOI":"10.1021\/ci300047k","volume":"52","author":"R Guha","year":"2012","unstructured":"Guha R (2012) Exploring uncharted territories: predicting activity cli\ufb00s in structure\u2013activity landscapes. J Chem Inf Model 52:2181\u20132191. https:\/\/doi.org\/10.1021\/ci300047k","journal-title":"J Chem Inf Model"},{"key":"1210_CR7","doi-asserted-by":"publisher","first-page":"2354","DOI":"10.1021\/ci300306a","volume":"52","author":"K Heikamp","year":"2012","unstructured":"Heikamp K, Hu X, Yan A, Bajorath J (2012) Prediction of activity cliffs using support vector machines. J Chem Inf Model 52:2354\u20132365. https:\/\/doi.org\/10.1021\/ci300306a","journal-title":"J Chem Inf Model"},{"key":"1210_CR8","doi-asserted-by":"publisher","first-page":"339","DOI":"10.1021\/ci900450m","volume":"50","author":"J Hussain","year":"2010","unstructured":"Hussain J, Rea C (2010) Computationally efficient algorithm to identify matched molecular pairs (MMPs) in large data sets. J Chem Inf Model 50:339\u2013348. https:\/\/doi.org\/10.1021\/ci900450m","journal-title":"J Chem Inf Model"},{"key":"1210_CR9","doi-asserted-by":"publisher","first-page":"1138","DOI":"10.1021\/ci3001138","volume":"52","author":"X Hu","year":"2012","unstructured":"Hu X, Hu Y, Vogt M et al (2012) MMP-Cliffs: systematic identification of activity cliffs on the basis of matched molecular pairs. J Chem Inf Model 52:1138\u20131145. https:\/\/doi.org\/10.1021\/ci3001138","journal-title":"J Chem Inf Model"},{"key":"1210_CR10","doi-asserted-by":"publisher","DOI":"10.1002\/minf.202000103","volume":"39","author":"S Tamura","year":"2020","unstructured":"Tamura S, Miyao T, Funatsu K (2020) Ligand\u2010based activity cliff prediction models with applicability domain. Mol Inform 39:2000103. https:\/\/doi.org\/10.1002\/minf.202000103","journal-title":"Mol Inform"},{"key":"1210_CR11","doi-asserted-by":"publisher","first-page":"4916","DOI":"10.3390\/molecules26164916","volume":"26","author":"S Tamura","year":"2021","unstructured":"Tamura S, Jasial S, Miyao T, Funatsu K (2021) Interpretation of ligand-based activity cliff prediction models using the matched molecular pair kernel. Molecules 26:4916. https:\/\/doi.org\/10.3390\/molecules26164916","journal-title":"Molecules"},{"key":"1210_CR12","doi-asserted-by":"publisher","first-page":"1631","DOI":"10.1021\/acs.jcim.6b00359","volume":"56","author":"D Horvath","year":"2016","unstructured":"Horvath D, Marcou G, Varnek A et al (2016) Prediction of activity cliffs using condensed graphs of reaction representations, descriptor recombination, support vector machine classification, and support vector regression. J Chem Inf Model 56:1631\u20131640. https:\/\/doi.org\/10.1021\/acs.jcim.6b00359","journal-title":"J Chem Inf Model"},{"key":"1210_CR13","doi-asserted-by":"publisher","first-page":"742","DOI":"10.1021\/ci100050t","volume":"50","author":"D Rogers","year":"2010","unstructured":"Rogers D, Hahn M (2010) Extended-connectivity fingerprints. J Chem Inf Model 50:742\u2013754. https:\/\/doi.org\/10.1021\/ci100050t","journal-title":"J Chem Inf Model"},{"key":"1210_CR14","doi-asserted-by":"publisher","first-page":"274","DOI":"10.1021\/acs.jcim.1c01163","volume":"62","author":"J Jim\u00e9nez-Luna","year":"2022","unstructured":"Jim\u00e9nez-Luna J, Skalic M, Weskamp N (2022) Benchmarking molecular feature attribution methods with activity cliffs. J Chem Inf Model 62:274\u2013283. https:\/\/doi.org\/10.1021\/acs.jcim.1c01163","journal-title":"J Chem Inf Model"},{"key":"1210_CR15","doi-asserted-by":"publisher","first-page":"5938","DOI":"10.1021\/acs.jcim.2c01073","volume":"62","author":"D van Tilborg","year":"2022","unstructured":"van Tilborg D, Alenicheva A, Grisoni F (2022) Exposing the limitations of molecular machine learning with activity cliffs. J Chem Inf Model 62:5938\u20135951. https:\/\/doi.org\/10.1021\/acs.jcim.2c01073","journal-title":"J Chem Inf Model"},{"key":"1210_CR16","doi-asserted-by":"publisher","first-page":"1157","DOI":"10.1007\/s10822-021-00380-y","volume":"35","author":"J Iqbal","year":"2021","unstructured":"Iqbal J, Vogt M, Bajorath J (2021) Prediction of activity cliffs on the basis of images using convolutional neural networks. J Comput Aided Mol Des 35:1157\u20131164. https:\/\/doi.org\/10.1007\/s10822-021-00380-y","journal-title":"J Comput Aided Mol Des"},{"key":"1210_CR17","doi-asserted-by":"publisher","first-page":"2341","DOI":"10.1021\/acs.jcim.2c00327","volume":"62","author":"J Park","year":"2022","unstructured":"Park J, Sung G, Lee S et al (2022) ACGCN: graph convolutional networks for activity cliff prediction between matched molecular pairs. J Chem Inf Model 62:2341\u20132351. https:\/\/doi.org\/10.1021\/acs.jcim.2c00327","journal-title":"J Chem Inf Model"},{"key":"1210_CR18","doi-asserted-by":"publisher","first-page":"898","DOI":"10.1039\/d2dd00077f","volume":"1","author":"H Chen","year":"2022","unstructured":"Chen H, Vogt M, Bajorath J (2022) DeepAC - conditional transformer-based chemical language model for the prediction of activity cliffs formed by bioactive compounds. Digit Discov 1:898\u2013909. https:\/\/doi.org\/10.1039\/d2dd00077f","journal-title":"Digit Discov"},{"key":"1210_CR19","doi-asserted-by":"publisher","first-page":"47","DOI":"10.1186\/s13321-023-00708-w","volume":"15","author":"M Dablander","year":"2023","unstructured":"Dablander M, Hanser T, Lambiotte R, Morris GM (2023) Exploring QSAR models for activity-cliff prediction. J Cheminform 15:47. https:\/\/doi.org\/10.1186\/s13321-023-00708-w","journal-title":"J Cheminform"},{"key":"1210_CR20","doi-asserted-by":"publisher","DOI":"10.1038\/s41467-023-41948-6","volume":"14","author":"J Deng","year":"2023","unstructured":"Deng J, Yang Z, Wang H, Ojima I, Samaras D, Wang F (2023) A systematic study of key elements underlying molecular property prediction. Nat Commun 14:6395. https:\/\/doi.org\/10.1038\/s41467-023-41948-6","journal-title":"Nat Commun"},{"key":"1210_CR21","doi-asserted-by":"publisher","DOI":"10.1038\/s41467-023-41967-3","volume":"14","author":"AL Dias","year":"2023","unstructured":"Dias AL, Bustillo L, Rodrigues T (2023) Limitations of representation learning in small molecule property prediction. Nat Commun 14:6394. https:\/\/doi.org\/10.1038\/s41467-023-41967-3","journal-title":"Nat Commun"},{"key":"1210_CR22","doi-asserted-by":"publisher","DOI":"10.1186\/s13321-022-00676-7","volume":"15","author":"S Tamura","year":"2023","unstructured":"Tamura S, Miyao T, Bajorath J (2023) Large-scale prediction of activity cliffs using machine and deep learning methods of increasing complexity. J Cheminform 15:4. https:\/\/doi.org\/10.1186\/s13321-022-00676-7","journal-title":"J Cheminform"},{"key":"1210_CR23","doi-asserted-by":"publisher","first-page":"1399","DOI":"10.1021\/cb3001028","volume":"7","author":"PM Petrone","year":"2012","unstructured":"Petrone PM, Simms B, Nigsch F et al (2012) Rethinking molecular similarity: comparing compounds on the basis of biological activity. ACS Chem Biol 7:1399\u20131409. https:\/\/doi.org\/10.1021\/cb3001028","journal-title":"ACS Chem Biol"},{"key":"1210_CR24","doi-asserted-by":"publisher","first-page":"1880","DOI":"10.1021\/ci500190p","volume":"54","author":"S Riniker","year":"2014","unstructured":"Riniker S, Wang Y, Jenkins JL, Landrum GA (2014) Using information from historical high-throughput screens to predict active compounds. J Chem Inf Model 54:1880\u20131891. https:\/\/doi.org\/10.1021\/ci500190p","journal-title":"J Chem Inf Model"},{"key":"1210_CR25","doi-asserted-by":"publisher","DOI":"10.1038\/s41467-023-37570-1","volume":"14","author":"N Moshkov","year":"2023","unstructured":"Moshkov N, Becker T, Yang K et al (2023) Predicting compound activity from phenotypic profiles and chemical structures. Nat Commun 14:1967. https:\/\/doi.org\/10.1038\/s41467-023-37570-1","journal-title":"Nat Commun"},{"key":"1210_CR26","doi-asserted-by":"publisher","first-page":"2695","DOI":"10.1021\/acs.jcim.3c01855","volume":"64","author":"WJ Godinez","year":"2024","unstructured":"Godinez WJ, Trifonov V, Fang B et al (2024) Compound activity prediction with dose-dependent transcriptomic profiles and deep learning. J Chem Inf Model 64:2695\u20132704. https:\/\/doi.org\/10.1021\/acs.jcim.3c01855","journal-title":"J Chem Inf Model"},{"key":"1210_CR27","doi-asserted-by":"publisher","DOI":"10.1038\/s41467-024-47171-1","volume":"15","author":"J Fredin Haslum","year":"2024","unstructured":"Fredin Haslum J, Lardeau CH, Karlsson J, Turkki R, Leuchowius KJ, Smith K, M\u00fcllers E (2024) Cell painting-based bioactivity prediction boosts high-throughput screening hit-rates and compound diversity. Nat Commun 15:3470. https:\/\/doi.org\/10.1038\/s41467-024-47171-1","journal-title":"Nat Commun"},{"key":"1210_CR28","doi-asserted-by":"publisher","first-page":"2944","DOI":"10.1021\/jm200026b","volume":"54","author":"M Wawer","year":"2011","unstructured":"Wawer M, Bajorath J (2011) Local structural changes, global data views: graphical substructure\u2212activity relationship trailing. J Med Chem 54:2944\u20132951. https:\/\/doi.org\/10.1021\/jm200026b","journal-title":"J Med Chem"},{"key":"1210_CR29","unstructured":"RDKit: Cheminformatics and Machine Learning Software. (2013) http:\/\/www.rdkit.org\/."},{"key":"1210_CR30","doi-asserted-by":"publisher","DOI":"10.4155\/fsoa-2018-0050","volume":"4","author":"M Vogt","year":"2018","unstructured":"Vogt M, Jasial S, Bajorath J (2018) Computationally derived compound profiling matrices. Future Sci OA 4:FSO327. https:\/\/doi.org\/10.4155\/fsoa-2018-0050","journal-title":"Future Sci OA"},{"key":"1210_CR31","doi-asserted-by":"publisher","first-page":"D955","DOI":"10.1093\/nar\/gkw1118","volume":"45","author":"Y Wang","year":"2017","unstructured":"Wang Y, Bryant SH, Cheng T et al (2017) PubChem BioAssay: 2017 update. Nucleic Acids Res 45:D955\u2013D963. https:\/\/doi.org\/10.1093\/nar\/gkw1118","journal-title":"Nucleic Acids Res"},{"key":"1210_CR32","doi-asserted-by":"publisher","DOI":"10.1186\/s13321-015-0069-3","author":"D Bajusz","year":"2015","unstructured":"Bajusz D, R\u00e1cz A, H\u00e9berger K (2015) Why is Tanimoto Index an appropriate choice for fingerprint-based similarity calculations? J Cheminform. https:\/\/doi.org\/10.1186\/s13321-015-0069-3","journal-title":"J Cheminform"},{"key":"1210_CR33","doi-asserted-by":"publisher","first-page":"5","DOI":"10.1023\/a:1010933404324","volume":"45","author":"L Breiman","year":"2001","unstructured":"Breiman L (2001) Random forests. Mach Learn 45:5\u201332. https:\/\/doi.org\/10.1023\/a:1010933404324","journal-title":"Mach Learn"},{"key":"1210_CR34","doi-asserted-by":"publisher","DOI":"10.1177\/15501329221106","volume":"18","author":"P Zhang","year":"2022","unstructured":"Zhang P, Jia Y, Shang Y (2022) Research and application of XGBoost in imbalanced data. Int J Distrib Sens Netw 18:15501329221106935. https:\/\/doi.org\/10.1177\/15501329221106","journal-title":"Int J Distrib Sens Netw"},{"key":"1210_CR35","doi-asserted-by":"publisher","unstructured":"Vapnik VN (2000) The nature of statistical learning theory.https:\/\/doi.org\/10.1007\/978-1-4757-3264-1","DOI":"10.1007\/978-1-4757-3264-1"},{"key":"1210_CR36","doi-asserted-by":"publisher","unstructured":"Brodersen KH, Ong CS, Stephan KE, Buhmann JM (2010) The Balanced Accuracy and Its Posterior Distribution. In: 2010 20th International Conference on Pattern Recognition, pp 3121\u20133124. https:\/\/doi.org\/10.1109\/ICPR.2010.764","DOI":"10.1109\/ICPR.2010.764"},{"key":"1210_CR37","doi-asserted-by":"publisher","first-page":"1397","DOI":"10.1093\/ije\/dyz274","volume":"49","author":"ACJ Janssens","year":"2020","unstructured":"Janssens ACJ, Martens FK (2020) Reflection on modern methods: revisiting the area under the ROC curve. Int J Epidemiol 49:1397\u20131403. https:\/\/doi.org\/10.1093\/ije\/dyz274","journal-title":"Int J Epidemiol"},{"key":"1210_CR38","doi-asserted-by":"publisher","first-page":"442","DOI":"10.1016\/0005-2795(75)90109-9","volume":"405","author":"BW Matthews","year":"1975","unstructured":"Matthews BW (1975) Comparison of the predicted and observed secondary structure of T4 phage lysozyme. Biochim Biophys Acta BBA Protein Struct 405:442\u2013451. https:\/\/doi.org\/10.1016\/0005-2795(75)90109-9","journal-title":"Biochim Biophys Acta BBA Protein Struct"},{"key":"1210_CR39","first-page":"431","volume":"26","author":"G Louppe","year":"2013","unstructured":"Louppe G, Wehenkel L, Sutera A, Geurts P (2013) Understanding variable importances in forests of randomized trees. Adv Neural Inf Process Syst 26:431\u2013439","journal-title":"Adv Neural Inf Process Syst"},{"key":"1210_CR40","first-page":"2825","volume":"12","author":"F Pedregosa","year":"2011","unstructured":"Pedregosa F, Varoquaux G, Gramfort A et al (2011) Scikit-learn: machine learning in Python. J Mach Learn Res 12:2825\u20132830","journal-title":"J Mach Learn Res"},{"key":"1210_CR41","doi-asserted-by":"publisher","first-page":"172","DOI":"10.4236\/ojs.2012.22019","volume":"2","author":"ICA Oyeka","year":"2012","unstructured":"Oyeka ICA, Ebuh GU (2012) Modified Wilcoxon signed-rank test. Open J Stat 2:172\u2013176. https:\/\/doi.org\/10.4236\/ojs.2012.22019","journal-title":"Open J Stat"}],"container-title":["Journal of Cheminformatics"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/link.springer.com\/article\/10.1186\/s13321-026-01210-9","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/link.springer.com\/content\/pdf\/10.1186\/s13321-026-01210-9.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/link.springer.com\/content\/pdf\/10.1186\/s13321-026-01210-9.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,6,3]],"date-time":"2026-06-03T10:31:15Z","timestamp":1780482675000},"score":1,"resource":{"primary":{"URL":"https:\/\/link.springer.com\/10.1186\/s13321-026-01210-9"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,6,3]]},"references-count":41,"journal-issue":{"issue":"1","published-online":{"date-parts":[[2026,12]]}},"alternative-id":["1210"],"URL":"https:\/\/doi.org\/10.1186\/s13321-026-01210-9","relation":{},"ISSN":["1758-2946"],"issn-type":[{"value":"1758-2946","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,6,3]]},"assertion":[{"value":"28 February 2026","order":1,"name":"received","label":"Received","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"25 April 2026","order":2,"name":"accepted","label":"Accepted","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"3 June 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":"The authors declare no competing interests.","order":2,"name":"Ethics","group":{"name":"EthicsHeading","label":"Competing interests"}}],"article-number":"72"}}