{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,5,13]],"date-time":"2026-05-13T06:19:39Z","timestamp":1778653179386,"version":"3.51.4"},"update-to":[{"DOI":"10.1371\/journal.pcbi.1013232","type":"new_version","label":"New version","source":"publisher","updated":{"date-parts":[[2026,4,29]],"date-time":"2026-04-29T00:00:00Z","timestamp":1777420800000}}],"reference-count":62,"publisher":"Public Library of Science (PLoS)","issue":"4","license":[{"start":{"date-parts":[[2026,4,22]],"date-time":"2026-04-22T00:00:00Z","timestamp":1776816000000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/publicdomain\/zero\/1.0\/"}],"funder":[{"DOI":"10.13039\/100007048","name":"Mayo Foundation for Medical Education and Research","doi-asserted-by":"publisher","id":[{"id":"10.13039\/100007048","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":["www.ploscompbiol.org"],"crossmark-restriction":false},"short-container-title":["PLoS Comput Biol"],"abstract":"<jats:p>Monotherapy cancer drug response prediction (DRP) models predict the response of a cell line to a given drug. Analyzing these models\u2019 performance includes assessing their ability to predict the response of cell lines to new drugs, i.e., drugs that are not in the training set. Drug-blind prediction displays greatly diminished performance or outright failure across a wide range of model architectures and different large pharmacogenomic datasets. Drug-blind failure is hypothesized to be caused by the relatively limited set of drugs present in these datasets. The time and cost associated with further cell line experiments is significant, and it is impossible to predict beforehand how much data would be enough to overcome drug-blind failure. We must first define how current data contributes to drug-blind failure before attempting to remedy drug-blind failure with further data collection. In this work, we quantify the extent to which drug-blind generalizability relies on mechanistic overlap of drugs between training and testing splits. We first identify that the majority of mixed set DRP model performance can be attributed to drug overfitting, likely inhibiting generalization and preventing accurate analysis. Then, by specifically probing the drug-blind ability of models, we reveal the sources of generalizable drug features are confined to shared mechanisms of action and related pathways. Furthermore, we observed that, for certain mechanisms, we can significantly improve performance by limiting the training of models to a single mechanism compared to training on all drugs simultaneously. Across multiple different model architectures examined in this paper, we observe that drug-blind performance is a poor benchmark for DRP as it does not describe model behavior, it describes dataset behavior. Our investigation displays that these deep learning models trained on large, monotherapy cell line panels can more accurately describe mechanism of action of drugs rather than their advertised connection to broader cancer biology.<\/jats:p>","DOI":"10.1371\/journal.pcbi.1013232","type":"journal-article","created":{"date-parts":[[2026,4,22]],"date-time":"2026-04-22T17:44:48Z","timestamp":1776879888000},"page":"e1013232","update-policy":"https:\/\/doi.org\/10.1371\/journal.pcbi.corrections_policy","source":"Crossref","is-referenced-by-count":1,"title":["Monotherapy cancer drug-blind response prediction is limited to intraclass generalization"],"prefix":"10.1371","volume":"22","author":[{"ORCID":"https:\/\/orcid.org\/0009-0006-5834-5353","authenticated-orcid":true,"given":"William G.","family":"Herbert","sequence":"first","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Nicholas","family":"Chia","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Paul A.","family":"Jensen","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Marina R. S.","family":"Walther-Antonio","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"340","published-online":{"date-parts":[[2026,4,22]]},"reference":[{"issue":"1","key":"pcbi.1013232.ref001","first-page":"7","article-title":"Cancer statistics, 2018","volume":"68","author":"RL Siegel","year":"2018","journal-title":"CA: a cancer journal for clinicians"},{"issue":"7","key":"pcbi.1013232.ref002","doi-asserted-by":"crossref","first-page":"1589","DOI":"10.1016\/j.cell.2024.02.009","article-title":"Embracing cancer complexity: Hallmarks of systemic disease","volume":"187","author":"C Swanton","year":"2024","journal-title":"Cell"},{"issue":"1","key":"pcbi.1013232.ref003","doi-asserted-by":"crossref","first-page":"360","DOI":"10.1093\/bib\/bbz171","article-title":"Deep learning for drug response prediction in cancer","volume":"22","author":"D Baptista","year":"2021","journal-title":"Brief Bioinform"},{"key":"pcbi.1013232.ref004","doi-asserted-by":"crossref","first-page":"1086097","DOI":"10.3389\/fmed.2023.1086097","article-title":"Deep learning methods for drug response prediction in cancer: Predominant and emerging trends","volume":"10","author":"A Partin","year":"2023","journal-title":"Front Med (Lausanne)"},{"key":"pcbi.1013232.ref005","doi-asserted-by":"crossref","unstructured":"Partin A, Vasanthakumari P, Narykov O, Wilke A, Koussa N, Jones SE, et al. Benchmarking community drug response prediction models: datasets, models, tools, and metrics for cross-dataset generalization analysis. arXiv preprint arXiv:250314356. 2025.","DOI":"10.1093\/bib\/bbaf667"},{"key":"pcbi.1013232.ref006","unstructured":"Overbeek JC, Partin A, Brettin TS, Chia N, Narykov O, Vasanthakumari P, et al. Assessing reusability of deep learning-based monotherapy drug response prediction models trained with omics data. 2024. https:\/\/arxiv.org\/abs\/240912215"},{"issue":"6","key":"pcbi.1013232.ref007","doi-asserted-by":"crossref","DOI":"10.1093\/bib\/bbab312","article-title":"A gentle introduction to understanding preclinical data for cancer pharmaco-omic modeling","volume":"22","author":"C Piyawajanusorn","year":"2021","journal-title":"Brief Bioinform"},{"issue":"1","key":"pcbi.1013232.ref008","doi-asserted-by":"crossref","first-page":"252","DOI":"10.1186\/s12859-021-04163-y","article-title":"Learning curves for drug response prediction in cancer cell lines","volume":"22","author":"A Partin","year":"2021","journal-title":"BMC Bioinformatics"},{"issue":"1","key":"pcbi.1013232.ref009","article-title":"A cross-study analysis of drug response prediction in cancer cell lines","volume":"23","author":"F Xia","year":"2022","journal-title":"Brief Bioinform"},{"issue":"7","key":"pcbi.1013232.ref010","doi-asserted-by":"crossref","first-page":"452","DOI":"10.1093\/jnci\/djt007","article-title":"The clinical relevance of cancer cell lines","volume":"105","author":"J-P Gillet","year":"2013","journal-title":"J Natl Cancer Inst"},{"key":"pcbi.1013232.ref011","doi-asserted-by":"crossref","unstructured":"Raji ID, Kumar IE, Horowitz A, Selbst A. The Fallacy of AI Functionality. In: 2022 ACM Conference on Fairness Accountability and Transparency, 2022. 959\u201372. https:\/\/doi.org\/10.1145\/3531146.3533158","DOI":"10.1145\/3531146.3533158"},{"key":"pcbi.1013232.ref012","doi-asserted-by":"crossref","unstructured":"Sambasivan N, Kapania S, Highfill H, Akrong D, Paritosh P, Aroyo LM. \u201cEveryone wants to do the model work, not the data work\u201d: Data Cascades in High-Stakes AI. In: Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems, 2021. 1\u201315. https:\/\/doi.org\/10.1145\/3411764.3445518","DOI":"10.1145\/3411764.3445518"},{"issue":"7480","key":"pcbi.1013232.ref013","doi-asserted-by":"crossref","first-page":"389","DOI":"10.1038\/nature12831","article-title":"Inconsistency in large pharmacogenomic studies","volume":"504","author":"B Haibe-Kains","year":"2013","journal-title":"Nature"},{"issue":"5","key":"pcbi.1013232.ref014","doi-asserted-by":"crossref","DOI":"10.1093\/bib\/bbae379","article-title":"Trust me if you can: a survey on reliability and interpretability of machine learning approaches for drug sensitivity prediction in cancer","volume":"25","author":"K Lenhof","year":"2024","journal-title":"Brief Bioinform"},{"key":"pcbi.1013232.ref015","doi-asserted-by":"crossref","first-page":"2333","DOI":"10.12688\/f1000research.9611.1","article-title":"Revisiting inconsistency in large pharmacogenomic studies","volume":"5","author":"Z Safikhani","year":"2016","journal-title":"F1000Res"},{"key":"pcbi.1013232.ref016","doi-asserted-by":"crossref","unstructured":"Gura T. Systems for identifying new drugs are often faulty. 1997.","DOI":"10.1126\/science.278.5340.1041"},{"key":"pcbi.1013232.ref017","doi-asserted-by":"crossref","first-page":"31619","DOI":"10.1038\/srep31619","article-title":"Multitask learning improves prediction of cancer drug sensitivity","volume":"6","author":"H Yuan","year":"2016","journal-title":"Sci Rep"},{"issue":"1","key":"pcbi.1013232.ref018","doi-asserted-by":"crossref","first-page":"11911","DOI":"10.1038\/s41598-023-39179-2","article-title":"A performance evaluation of drug response prediction models for individual drugs","volume":"13","author":"A Park","year":"2023","journal-title":"Sci Rep"},{"key":"pcbi.1013232.ref019","doi-asserted-by":"crossref","DOI":"10.1093\/bioinformatics\/btaf255","article-title":"Understanding the sources of performance in deep drug response models reveals insights and improvements","volume":"41","author":"N Branson","year":"2025","journal-title":"Bioinformatics"},{"issue":"1","key":"pcbi.1013232.ref020","doi-asserted-by":"crossref","first-page":"33","DOI":"10.1186\/s13321-025-00972-y","article-title":"The specification game: rethinking the evaluation of drug response prediction for precision oncology","volume":"17","author":"F Codic\u00e8","year":"2025","journal-title":"J Cheminform"},{"issue":"2","key":"pcbi.1013232.ref021","doi-asserted-by":"crossref","DOI":"10.1093\/bib\/bbaf134","article-title":"Data imbalance in drug response prediction: multi-objective optimization approach in deep learning setting","volume":"26","author":"O Narykov","year":"2025","journal-title":"Brief Bioinform"},{"key":"pcbi.1013232.ref022","unstructured":"Tan J, Yang J, Wu S, Chen G, Zhao J. A critical look at the current train\/test split in machine learning. arXiv preprint. 2021. https:\/\/arxiv.org\/abs\/2106.04525"},{"issue":"10","key":"pcbi.1013232.ref023","doi-asserted-by":"crossref","first-page":"2570","DOI":"10.1021\/ci300338w","article-title":"Does rational selection of training and test sets improve the outcome of QSAR modeling?","volume":"52","author":"TM Martin","year":"2012","journal-title":"Journal of Chemical Information and Modeling"},{"issue":"8","key":"pcbi.1013232.ref024","doi-asserted-by":"crossref","first-page":"1481","DOI":"10.1038\/s41592-024-02305-7","article-title":"Large-scale foundation model on single-cell transcriptomics","volume":"21","author":"M Hao","year":"2024","journal-title":"Nat Methods"},{"issue":"1","key":"pcbi.1013232.ref025","article-title":"GraphCDR: a graph neural network method with contrastive learning for cancer drug response prediction","volume":"23","author":"X Liu","year":"2022","journal-title":"Brief Bioinform"},{"issue":"6","key":"pcbi.1013232.ref026","article-title":"Interpretable deep learning architectures for improving drug response prediction performance: myth or reality?","volume":"39","author":"Y Li","year":"2023","journal-title":"Bioinformatics"},{"issue":"3","key":"pcbi.1013232.ref027","doi-asserted-by":"crossref","first-page":"135","DOI":"10.18632\/oncotarget.240","article-title":"Ras\/Raf\/MEK\/ERK and PI3K\/PTEN\/Akt\/mTOR inhibitors: rationale and importance to inhibiting these pathways in human health","volume":"2","author":"WH Chappell","year":"2011","journal-title":"Oncotarget"},{"issue":"9","key":"pcbi.1013232.ref028","doi-asserted-by":"crossref","first-page":"954","DOI":"10.18632\/oncotarget.652","article-title":"Mutations and deregulation of Ras\/Raf\/MEK\/ERK and PI3K\/PTEN\/Akt\/mTOR cascades which alter therapy response","volume":"3","author":"JA McCubrey","year":"2012","journal-title":"Oncotarget"},{"issue":"5","key":"pcbi.1013232.ref029","doi-asserted-by":"crossref","first-page":"1021","DOI":"10.1158\/1535-7163.MCT-13-0639","article-title":"Picking the point of inhibition: a comparative review of PI3K\/AKT\/mTOR pathway inhibitors","volume":"13","author":"R Dienstmann","year":"2014","journal-title":"Mol Cancer Ther"},{"issue":"6367","key":"pcbi.1013232.ref030","doi-asserted-by":"crossref","DOI":"10.1126\/science.aan4368","article-title":"The target landscape of clinical kinase drugs","volume":"358","author":"S Klaeger","year":"2017","journal-title":"Science"},{"issue":"8","key":"pcbi.1013232.ref031","doi-asserted-by":"crossref","first-page":"2248","DOI":"10.1038\/bjc.2013.550","article-title":"The FGFR1 inhibitor PD173074 induces mesenchymal-epithelial transition through the transcription factor AP-1","volume":"109","author":"PT Nguyen","year":"2013","journal-title":"Br J Cancer"},{"issue":"41","key":"pcbi.1013232.ref032","doi-asserted-by":"crossref","first-page":"5486","DOI":"10.1038\/onc.2008.244","article-title":"Class I PI3K in oncogenic cellular transformation","volume":"27","author":"L Zhao","year":"2008","journal-title":"Oncogene"},{"issue":"10","key":"pcbi.1013232.ref033","doi-asserted-by":"crossref","first-page":"741","DOI":"10.1038\/s41573-021-00209-1","article-title":"PI3K inhibitors are finally coming of age","volume":"20","author":"B Vanhaesebroeck","year":"2021","journal-title":"Nat Rev Drug Discov"},{"issue":"14","key":"pcbi.1013232.ref034","doi-asserted-by":"crossref","DOI":"10.1093\/bioinformatics\/btz318","article-title":"MOLI: multi-omics late integration with deep neural networks for drug response prediction","volume":"35","author":"H Sharifi-Noghabi","year":"2019","journal-title":"Bioinformatics"},{"key":"pcbi.1013232.ref035","first-page":"2024","article-title":"CancerFoundation: A single-cell RNA sequencing foundation model to decipher drug resistance in cancer","author":"A Theus","year":"2024","journal-title":"bioRxiv"},{"issue":"1","key":"pcbi.1013232.ref036","doi-asserted-by":"crossref","first-page":"179","DOI":"10.1038\/s41598-024-83090-3","article-title":"Drug discovery and mechanism prediction with explainable graph neural networks","volume":"15","author":"C Wang","year":"2025","journal-title":"Sci Rep"},{"issue":"7","key":"pcbi.1013232.ref037","doi-asserted-by":"crossref","first-page":"912","DOI":"10.1080\/09553002.2019.1589653","article-title":"The Warburg effect: essential part of metabolic reprogramming and central contributor to cancer progression","volume":"95","author":"P Vaupel","year":"2019","journal-title":"Int J Radiat Biol"},{"issue":"5081","key":"pcbi.1013232.ref038","doi-asserted-by":"crossref","first-page":"447","DOI":"10.1126\/science.1411538","article-title":"Neural computing in cancer drug development: predicting mechanism of action","volume":"258","author":"JN Weinstein","year":"1992","journal-title":"Science"},{"issue":"5","key":"pcbi.1013232.ref039","article-title":"AdaSemb: an adaptive knowledge-driven deep learning framework integrating cancer protein assemblies for predicting PI3K\u03b1 inhibitor response and resistance","volume":"26","author":"Z Li","year":"2025","journal-title":"Brief Bioinform"},{"issue":"12","key":"pcbi.1013232.ref040","doi-asserted-by":"crossref","DOI":"10.1001\/jamanetworkopen.2021.38793","article-title":"Anticancer Drugs Approved by the US Food and Drug Administration From 2009 to 2020 According to Their Mechanism of Action","volume":"4","author":"T Olivier","year":"2021","journal-title":"JAMA Netw Open"},{"key":"pcbi.1013232.ref041","first-page":"18661","article-title":"Supervised contrastive learning","volume":"33","author":"P Khosla","year":"2020","journal-title":"Advances in Neural Information Processing Systems"},{"key":"pcbi.1013232.ref042","unstructured":"Graf F, Hofer C, Niethammer M, Kwitt R. Dissecting supervised contrastive learning. In: Proceedings of the International Conference on Machine Learning, 2021. 3821\u201330."},{"issue":"1","key":"pcbi.1013232.ref043","doi-asserted-by":"crossref","first-page":"12303","DOI":"10.1038\/s41598-024-62956-6","article-title":"Reliable anti-cancer drug sensitivity prediction and prioritization","volume":"14","author":"K Lenhof","year":"2024","journal-title":"Sci Rep"},{"issue":"2","key":"pcbi.1013232.ref044","doi-asserted-by":"crossref","first-page":"461","DOI":"10.1093\/bioinformatics\/btab650","article-title":"TGSA: protein-protein association-based twin graph neural networks for drug response prediction with similarity augmentation","volume":"38","author":"Y Zhu","year":"2022","journal-title":"Bioinformatics"},{"issue":"1","key":"pcbi.1013232.ref045","doi-asserted-by":"crossref","first-page":"146","DOI":"10.1109\/TCBB.2021.3060430","article-title":"Graph Convolutional Networks for Drug Response Prediction","volume":"19","author":"T Nguyen","year":"2022","journal-title":"IEEE\/ACM Trans Comput Biol Bioinform"},{"key":"pcbi.1013232.ref046","article-title":"DeepCDR: a hybrid graph convolutional network for predicting cancer drug response","volume":"36","author":"Q Liu","year":"2020","journal-title":"Bioinformatics"},{"issue":"3","key":"pcbi.1013232.ref047","doi-asserted-by":"crossref","DOI":"10.1093\/bib\/bbac100","article-title":"DeepTTA: a transformer-based model for predicting cancer drug response","volume":"23","author":"L Jiang","year":"2022","journal-title":"Brief Bioinform"},{"issue":"1","key":"pcbi.1013232.ref048","doi-asserted-by":"crossref","first-page":"408","DOI":"10.1186\/s12859-019-2910-6","article-title":"Improving prediction of phenotypic drug response on cancer cell lines using deep convolutional network","volume":"20","author":"P Liu","year":"2019","journal-title":"BMC Bioinformatics"},{"issue":"49","key":"pcbi.1013232.ref049","doi-asserted-by":"crossref","DOI":"10.1073\/pnas.2208458119","article-title":"High-throughput functional annotation of natural products by integrated activity profiling","volume":"119","author":"SK Hight","year":"2022","journal-title":"Proc Natl Acad Sci U S A"},{"issue":"2","key":"pcbi.1013232.ref050","article-title":"MAST1 Drives Cisplatin Resistance in Human Cancers by Rewiring cRaf-Independent MEK Activation","volume":"34","author":"L Jin","year":"2018","journal-title":"Cancer Cell"},{"issue":"1","key":"pcbi.1013232.ref051","doi-asserted-by":"crossref","DOI":"10.1093\/biomethods\/bpaf033","article-title":"Optimizing drug synergy prediction through categorical embeddings in deep neural networks","volume":"10","author":"M Gonz\u00e1lez Lastre","year":"2025","journal-title":"Biol Methods Protoc"},{"issue":"3","key":"pcbi.1013232.ref052","doi-asserted-by":"crossref","DOI":"10.1093\/bib\/bbac075","article-title":"A review of machine learning approaches for drug synergy prediction in cancer","volume":"23","author":"A Torkamannia","year":"2022","journal-title":"Brief Bioinform"},{"key":"pcbi.1013232.ref053","unstructured":"U S Food and Drug Administration. Roadmap to Reducing Animal Testing in Preclinical Safety Studies. 2025."},{"key":"pcbi.1013232.ref054","unstructured":"DepMap B. DepMap 24Q2 Public. 2024. https:\/\/doi.org\/10.25452\/figshare.plus.27993248.v1"},{"issue":"10","key":"pcbi.1013232.ref055","doi-asserted-by":"crossref","first-page":"1113","DOI":"10.1038\/ng.2764","article-title":"The Cancer Genome Atlas Pan-Cancer analysis project","volume":"45","author":"Cancer Genome Atlas Research Network","year":"2013","journal-title":"Nat Genet"},{"issue":"7603","key":"pcbi.1013232.ref056","doi-asserted-by":"crossref","first-page":"333","DOI":"10.1038\/nature17987","article-title":"Reproducible pharmacogenomic profiling of cancer cell line panels","volume":"533","author":"PM Haverty","year":"2016","journal-title":"Nature"},{"issue":"2","key":"pcbi.1013232.ref057","doi-asserted-by":"crossref","first-page":"109","DOI":"10.1038\/nchembio.1986","article-title":"Correlating chemical sensitivity and basal gene expression reveals mechanism of action","volume":"12","author":"MG Rees","year":"2016","journal-title":"Nat Chem Biol"},{"issue":"11","key":"pcbi.1013232.ref058","doi-asserted-by":"crossref","first-page":"1210","DOI":"10.1158\/2159-8290.CD-15-0235","article-title":"Harnessing Connectivity in a Large-Scale Small-Molecule Sensitivity Dataset","volume":"5","author":"B Seashore-Ludlow","year":"2015","journal-title":"Cancer Discov"},{"key":"pcbi.1013232.ref059","article-title":"Genomics of Drug Sensitivity in Cancer (GDSC): a resource for therapeutic biomarker discovery in cancer cells","volume":"41","author":"W Yang","year":"2013","journal-title":"Nucleic Acids Res"},{"issue":"3","key":"pcbi.1013232.ref060","doi-asserted-by":"crossref","first-page":"740","DOI":"10.1016\/j.cell.2016.06.017","article-title":"A Landscape of Pharmacogenomic Interactions in Cancer","volume":"166","author":"F Iorio","year":"2016","journal-title":"Cell"},{"key":"pcbi.1013232.ref061","doi-asserted-by":"crossref","first-page":"486","DOI":"10.1186\/s12859-018-2509-3","article-title":"Predicting tumor cell line response to drug pairs with deep learning","volume":"19","author":"F Xia","year":"2018","journal-title":"BMC Bioinformatics"},{"issue":"2","key":"pcbi.1013232.ref062","doi-asserted-by":"crossref","first-page":"301","DOI":"10.1111\/j.1467-9868.2005.00503.x","article-title":"Regularization and Variable Selection Via the Elastic Net","volume":"67","author":"H Zou","year":"2005","journal-title":"Journal of the Royal Statistical Society Series B: Statistical Methodology"}],"updated-by":[{"DOI":"10.1371\/journal.pcbi.1013232","type":"new_version","label":"New version","source":"publisher","updated":{"date-parts":[[2026,4,29]],"date-time":"2026-04-29T00:00:00Z","timestamp":1777420800000}}],"container-title":["PLOS Computational Biology"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/dx.plos.org\/10.1371\/journal.pcbi.1013232","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,4,29]],"date-time":"2026-04-29T18:05:14Z","timestamp":1777485914000},"score":1,"resource":{"primary":{"URL":"https:\/\/dx.plos.org\/10.1371\/journal.pcbi.1013232"}},"subtitle":[],"editor":[{"given":"Hailin","family":"Chen","sequence":"first","affiliation":[],"role":[{"role":"editor","vocabulary":"crossref"}]}],"short-title":[],"issued":{"date-parts":[[2026,4,22]]},"references-count":62,"journal-issue":{"issue":"4","published-online":{"date-parts":[[2026,4,22]]}},"URL":"https:\/\/doi.org\/10.1371\/journal.pcbi.1013232","relation":{},"ISSN":["1553-7358"],"issn-type":[{"value":"1553-7358","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,4,22]]}}}