Abstract
Potency (IC50) prediction of small molecules is pivotal for anticancer drug development. This study benchmarked five deep learning (DL) models for IC50 prediction—DeepCDR, DrugCell, PaccMann, Precily, and tCNN—against a simple mean-based Baseline using standardized GDSC datasets and recently published anticancer compounds. To ensure practicality, conventional error metrics were supplemented with percentage error, log error, three-sigma limit, and a newly proposed Experimental Variability-Aware Prediction Accuracy statistic. The models performed well on randomly split data and unseen cell lines but showed sharply reduced accuracy for unseen compounds. Though all DL models exhibited similar performance trends, DeepCDR, DrugCell, and tCNN held a slight edge in most testing scenarios. Interestingly, several DL algorithms could not significantly outperform the Baseline model in many tests. Assessing prediction error against physicochemical and biological properties of compounds and cell lines revealed weak correlation, highlighting an underexplored aspect of model performance. A user-friendly web server (https://nlplab1.isical.ac.in/ic50.php) was also developed for IC50 prediction of new compounds against cancer cell lines.
Similar content being viewed by others
Data availability
The data supporting the findings of this study are available within the article and its Supplementary Information file. Supplementary Data 1 contains data related to EVAPA derivation, unseen anticancer compounds from literature, and five-fold cross-validation results (only prediction data from the fold with the best EVAPA for each model and split included). The numerical source data of main figures are available in Supplementary Data 2. All other data are available from the corresponding author upon reasonable request.
Code availability
This work mainly uses author-published code for the DL models.
References
Bray, F. et al. Global cancer statistics 2022: Globocan estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J. Clin. 74, 229–263 (2024).
Fatima, I. et al. Breakthroughs in ai and multi-omics for cancer drug discovery: a review. Eur. J. Med. Chem. 280, 116925 (2024).
Zhang, L., Shi, L., Soars, S. M., Kamps, J. & Yin, H. Discovery of novel small-molecule inhibitors of nf-κb signaling with antiinflammatory and anticancer properties. J. Med. Chem. 61, 5881–5899 (2018).
Michelini, E., Cevenini, L., Mezzanotte, L., Coppa, A. & Roda, A. Cell-based assays: fuelling drug discovery. Anal. Bioanal. Chem. 398, 227–238 (2010).
Lee, J. A. & Berg, E. L. Neoclassic drug discovery: the case for lead generation using phenotypic and functional approaches. J. Biomol. Screen. 18, 1143–1155 (2013).
Cava, C. & Castiglioni, I. Integration of molecular docking and in vitro studies: a powerful approach for drug discovery in breast cancer. Appl. Sci. 10, 6981 (2020).
Stevens, J. A. et al. Molecular dynamics simulation of an entire cell. Front. Chem. 11, 1106495 (2023).
Smith, J. S., Roitberg, A. E. & Isayev, O. Transforming computational drug discovery with machine learning and AI. ACS Med. Chem. Lett. 9, 1065–1069 (2018).
Tropsha, A. Best practices for qsar model development, validation, and exploitation. Mol. Inform. 29, 476–488 (2010).
https://www.cancerrxgene.org/compound/Foretinib/2040/overview/ic50 (2025).
Zhu, H. Big data and artificial intelligence modeling for drug discovery. Annu. Rev. of Pharmacol. Toxicol. 60, 573–589 (2020).
Cai, C. et al. Transfer learning for drug discovery. J. Med. Chem. 63, 8683–8694 (2020).
Badwan, B. A. et al. Machine learning approaches to predict drug efficacy and toxicity in oncology. Cell Rep. Methods 3, 100413 (2023).
Hinton, G. E. & Salakhutdinov, R. R. Reducing the dimensionality of data with neural networks. Science 313, 504–507 (2006).
LeCun, Y., Bengio, Y. & Hinton, G. Deep learning. Nature 521, 436–444 (2015).
Yang, W. et al. Genomics of drug sensitivity in cancer (gdsc): a resource for therapeutic biomarker discovery in cancer cells. Nucleic. Acids Res. 41, D955–D961 (2012).
Barretina, J. et al. The cancer cell line encyclopedia enables predictive modelling of anticancer drug sensitivity. Nature 483, 603–607 (2012).
https://www.cancer.gov/tcga. The Cancer Genome Atlas (TCGA).
Chen, J. & Zhang, L. A survey and systematic assessment of computational methods for drug response prediction. Brief. Bioinformatics 22, 232–246 (2021).
Park, A. et al. A comprehensive evaluation of regression-based drug responsiveness prediction models, using cell viability inhibitory concentrations (IC50 values). Bioinformatics 38, 2810–2817 (2022).
Costello, J. C. et al. A community effort to assess and improve drug sensitivity prediction algorithms. Nat. Biotechnol. 32, 1202–1212 (2014).
Liu, Q., Hu, Z., Jiang, R. & Zhou, M. Deepcdr: a hybrid graph convolutional network for predicting cancer drug response. Bioinformatics 36, i911–i918 (2020).
Kuenzi, B. M. et al. Predicting drug response and synergy using a deep learning model of human cancer cells. Cancer Cell 38, 672–684 (2020).
Oskooei, A. et al. PaccMann: prediction of anticancer compound sensitivity with multi-modal attention-based neural networks. In Proc. Workshop on Machine Learning for Molecules and Materials, 32nd Conference on Neural Information Processing Systems (NIPS, 2018).
Chawla, S. et al. Gene expression based inference of cancer drug sensitivity. Nat. Commun. 13, 5680 (2022).
Liu, P., Li, H., Li, S. & Leung, K.-S. Improving prediction of phenotypic drug response on cancer cell lines using deep convolutional network. BMC Bioinformatics 20, 1–14 (2019).
Damiani, E., Solorio, J. A., Doyle, A. P. & Wallace, H. M. How reliable are in vitro IC50 values? values vary with cytotoxicity assays in human glioblastoma cells. Toxicol. Lett. 302, 28–34 (2019).
He, Y. et al. The changing 50% inhibitory concentration (IC50) of cisplatin: a pilot study on the artifacts of the mtt assay and the precise measurement of density-dependent chemoresistance in ovarian cancer. Oncotarget 7, 70803 (2016).
Brooks, E. A. et al. Applicability of drug response metrics for cancer studies using biomaterials. Philos. Trans. R. Soc. Lond. B Biol. Sci. 374, 20180226 (2019).
Li, Y. et al. Discovery, synthesis, and evaluation of novel dual inhibitors of a vascular endothelial growth factor receptor and poly (adp-ribose) polymerase for BRCA wild-type breast cancer therapy. J. Med. Chem. 66, 12069–12100 (2023).
Zhang, L. et al. Design, synthesis, and anti-cancer evaluation of novel cyclic phosphate prodrug of gemcitabine. J. Med. Chem. 66, 4150–4166 (2023).
Wang, Y., Lv, Z., Chen, F., Wang, X. & Gou, S. Conjugates derived from lapatinib derivatives with cancer cell stemness inhibitors effectively reversed drug resistance in triple-negative breast cancer. J. Med. Chem. 64, 12877–12892 (2021).
Tang, Z. et al. Novel covalent probe selectively targeting glutathione peroxidase 4 in vivo: Potential applications in pancreatic cancer therapy. J. Med. Chem. 67, 1872–1887 (2024).
Gu, H. et al. Discovery of a highly potent and selective myof inhibitor with improved water solubility for the treatment of gastric cancer. J. Med. Chem. 66, 16917–16938 (2023).
Fang, W. et al. Discovery of the first-in-class rorγ covalent inhibitors for treatment of castration-resistant prostate cancer. J. Med. Chem. 67, 1481–1499 (2024).
Li, Y. et al. Synthesis and anti-tumor activity of nitrogen-containing derivatives of the natural product diphyllin. Eur. J. Med. Chem. 243, 114708 (2022).
Litvinova, V. A. et al. Naphthoindole-2-carboxamides as a lipophilic chemotype of hetarene-anthraquinones potent against p-gp resistant tumor cells. Eur. J. Med. Chem. 281, 117013 (2025).
Li, S.-S. et al. Design, synthesis, and biological evaluation of novel benzimidazole derivatives as anti-cervical cancer agents through PI3K/Akt/mTOR pathway and tubulin inhibition. Eur. J. Med. Chem. 271, 116425 (2024).
Borude, A. S., Deshmukh, S. R., Tiwari, S. V., Kumar, S. H. & Thopate, S. R. Design and synthesis of novel thiazolo[5,4-b]pyridine derivatives as potent and selective egfr-tk inhibitors targeting resistance mutations in non-small cell lung cancer. Eur. J. Med. Chem. 276, 116727 (2024).
Benjamin, D. J. et al. Redefine statistical significance. Nat. Hum. Behav. 2, 6–10 (2018).
Shi, Y., Xu, W. & Hu, P. Out of distribution learning in bioinformatics: advancements and challenges. Brief. Bioinformatics 26, bbaf294 (2025).
Pirie, R. et al. An analysis of the physicochemical properties of oral drugs from 2000 to 2022. RSC. Med. Chem. 15, 3125–3132 (2024).
Sterling, T. & Irwin, J. J. Zinc 15–ligand discovery for everyone. J. Chem. Inf. Model. 55, 2324–2337 (2015).
Scarpino, A., Ferenczy, G. G. & Keseru, G. M. Comparative evaluation of covalent docking tools. J. Chem. Inf. Model. 58, 1441–1458 (2018).
Pukelsheim, F. The three sigma rule. Am. Stat. 48, 88–91 (1994).
Johnson, S. R. et al. A simple strategy for mitigating the effect of data variability on the identification of active chemotypes from high-throughput screening data. SLAS Discov. 12, 276–284 (2007).
Niepel, M. et al. A multi-center study on the reproducibility of drug-response assays in mammalian cell lines. Cell Syst. 9, 35–48.e5 (2019).
Haibe-Kains, B. et al. Inconsistency in large pharmacogenomic studies. Nature 504, 389–393 (2013).
Kalliokoski, T., Kramer, C., Vulpetti, A. & Gedeck, P. Comparability of mixed IC50 data–a statistical analysis. PLoS ONE 8, e61007 (2013).
Das, K. R. & Imon, A. A brief review of tests for normality. Am. J. Theor. Appl. Stat. 5, 5–12 (2016).
Gordon, S. The Normal Distribution 1 edn (Sydney: Mathematics Learning Centre, University of Sydney, 2006)
Mikolov, T., Chen, K., Corrado, G. & Dean, J. Distributed representations of words and phrases and their compositionality. In Proc. 26th Annual Conference on Neural Information Processing Systems (NIPS) (Curran Associates Inc., 2013).
Nguyen, E. et al. Hyenadna: long-range genomic sequence modeling at single nucleotide resolution. In Proc. Advances in Neural Information Processing Systems, Oh, A. et al. (eds) 43177–43201 https://proceedings.neurips.cc/paper_files/paper/2023/file/86ab6927ee4ae9bde4247793c46797c7-Paper-Conference.pdf (Curran Associates, Inc., 2023).
Acknowledgements
The authors sincerely thank the Master’s students at the Indian Statistical Institute—Siva Kumar Lakkoju, Sneha Tiwari, and Abhishek Bale—for their coding and testing contributions to this research. We are grateful to Debasrija Mondal for assistance with the illustration of Fig. 1. We appreciate and express gratitude for the comments and suggestions of the anonymous reviewers, which have greatly helped improve the work.
Author information
Authors and Affiliations
Contributions
The first author U.Garai conceived the concept, designed the experiments, collated data, provided domain knowledge, analysed the results, critically reviewed the research, and took the lead in writing the manuscript. A.S.P. managed all computations, implemented the deep learning models, and developed the web server. K.G. assisted with computations, managed run-wise results, and contributed to the analysis. D.B.S. provided domain-specific support and critically reviewed the research from the perspective of medicinal chemistry. The corresponding author, U.Garain supervised and managed the entire research, including the experiments and manuscript preparation. The first two authors, U.Garai and A.S.P., contributed equally to this work.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Peer review
Peer review information
Communications Chemistry thanks Martin Vogt and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
About this article
Cite this article
Garai, U., Pal, A.S., Ghosh, K. et al. Benchmarking deep learning models for predicting anticancer drug potency (IC50) with insights for medicinal chemists. Commun Chem (2026). https://doi.org/10.1038/s42004-026-01916-9
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s42004-026-01916-9
