Structure–function conditioned molecular generation with PocketGPT enables prospective discovery of PARP1 inhibitors

Rong Qian; Jilong Duan; Xiaoping Liu; Wenxuan Wang; Tingfei Zhu; Dejun Jiang; Jin Liu; Youchao Deng; Aiping Lyu; Dongsheng Cao

doi:10.59717/j.xinn-drugdisc.2026.100028

The Innovation Drug Discovery > 2026 Vol. 1 > No. 2 > 100028

Next Article Previous Article

ARTICLE Open Access

Cite PDF

Structure–function conditioned molecular generation with PocketGPT enables prospective discovery of PARP1 inhibitors

1.
Xiangya School of Pharmaceutical Sciences, Central South University, Changsha 410013, Hunan, China
2.
School of Chinese Medicine, Hong Kong Baptist University, Hong Kong SAR 999077, China
3.
Institute of systems medicine and health sciences, Hong Kong Baptist University, Hong Kong SAR 999077, China
4.
These authors contributed equally to this work

More Information

Corresponding authors: dengyouchao1206@163.com (Y.D.); aipinglu@hkbu.edu.hk (A.L.); oriental-cds@163.com (D.C.)
Received Date: March 30, 2026

Accepted Date: July 07, 2026

Published Online: July 15, 2026

The Innovation Drug Discovery 1(2), https://doi.org/10.59717/j.xinn-drugdisc.2026.100028 | Cite this article

GRAPHICAL ABSTRACT

GRAPHICAL ABSTRACT

DownLoad: Full size image

PUBLIC SUMMARY

PUBLIC SUMMARY
1. PocketGPT uses protein pocket and functional information to guide controllable molecule design.
  
  It integrates pocket context, molecular properties, fragments, and interactions to enrich promising candidates.
  
  A prospective poly(ADP-ribose) polymerase 1 (PARP1) study yielded an antitumor nanomolar inhibitor.

ABSTRACT

ABSTRACT

Structure-based molecular generation has become increasingly prominent in drug discovery; however, many generated binders still fail to satisfy the physicochemical and functional constraints required for pharmacological viability. Here, we present PocketGPT, a pocket-conditioned generative framework that demonstrates structure-aware molecular generation can be effectively enabled without explicit three-dimensional ligand construction, by jointly conditioning protein pocket context and medicinal-chemistry constraints during generation. By co-conditioning physicochemical property targets, fragment priors, and protein–ligand interaction fingerprints, PocketGPT enables structure–function co-generative control, allowing directional modulation of drug-like properties while partially retaining binding-relevant interaction patterns across diverse targets. In a prospective design–make–test campaign targeting poly(ADP-ribose) polymerase 1 (PARP1), PocketGPT identified a novel inhibitor with nanomolar binding affinity (K_d = 4.88 nM) and favorable tolerability in mice. Together, these results demonstrate that joint structural and pharmacological conditioning, embedded directly in the generative process, constitutes a viable and controllable representation for structure-aware molecular generation, offering a complementary pathway to geometry-first approaches for producing experimentally actionable chemical matter.

FullText(HTML)

REFERENCES

[1]	Dickson, M., and Gagnon, J.P. (2004). Key factors in the rising cost of new drug discovery and development. Nat Rev Drug Discov 3:417−429. DOI:10.1038/nrd1382 View in Article CrossRef Google Scholar
[2]	Paul, S.M., Mytelka, D.S., Dunwiddie, C.T., et al. (2010). How to improve R&D productivity: the pharmaceutical industry's grand challenge. Nat Rev Drug Discov 9:203−214. DOI:10.1038/nrd3078 View in Article CrossRef Google Scholar
[3]	Sun, D., Gao, W., Hu, H., and Zhou, S. (2022). Why 90% of clinical drug development fails and how to improve it. Acta Pharm Sin B 12(7):3049−3062. DOI:10.1016/j.apsb.2022.02.002 View in Article CrossRef Google Scholar
[4]	Vamathevan, J., Clark, D., Czodrowski, P., et al. (2019). Applications of machine learning in drug discovery and development. Nat Rev Drug Discov 18:463−477. DOI:10.1038/s41573-019-0024-5 View in Article CrossRef Google Scholar
[5]	Catacutan, D.B., Alexander, J., Arnold, A., and Stokes, J.M. (2024). Machine learning in preclinical drug discovery. Nat Chem Biol 20:960−973. DOI:10.1038/s41589-024-01679-1 View in Article CrossRef Google Scholar
[6]	Du, Y., Jamasb, A.R., Guo, J., et al. (2024). Machine learning-aided generative molecular design. Nat Mach Intell 6:589−604. DOI:10.1038/s42256-024-00843-5 View in Article CrossRef Google Scholar
[7]	Vert, J.-P. (2023). How will generative AI disrupt data science in drug discovery. Nat Biotechnol 41:750−751. DOI:10.1038/s41587-023-01789-6 View in Article CrossRef Google Scholar
[8]	Khater, T., Alkhatib, S.A., AlShehhi, A., et al. (2025). Generative artificial intelligence based models optimization towards molecule design enhancement. J Cheminformatics 17:116. DOI:10.1186/s13321-025-01059-4 View in Article CrossRef Google Scholar
[9]	Tong, X., Liu, X., Tan, X., et al. (2021). Generative Models for De Novo Drug Design. J Med Chem 64:14011−14027. DOI:10.1021/acs.jmedchem.1c00927 View in Article CrossRef Google Scholar
[10]	Wang, M., Wang, Z., Sun, H., et al. (2022). Deep learning approaches for de novo drug design: An overview. Curr Opin Struc Biol 72:135−144. DOI:10.1016/j.sbi.2021.10.001 View in Article CrossRef Google Scholar
[11]	Ivanenkov, Y., Zagribelnyy, B., Malyshev, A., et al. (2023). The Hitchhiker’s Guide to Deep Learning Driven Generative Chemistry. Acs Med Chem Lett 14:901−915. DOI:10.1021/acsmedchemlett.3c00041 View in Article CrossRef Google Scholar
[12]	Deng, J., Yang, Z., Ojima, I., et al. (2022). Artificial intelligence in drug discovery: applications and techniques. Brief Bioinform 23:bbab430. DOI:10.1093/bib/bbab430 View in Article CrossRef Google Scholar
[13]	Tropsha, A., Isayev, O., Varnek, A., et al. (2024). Integrating QSAR modelling and deep learning in drug discovery: the emergence of deep QSAR. Nat Rev Drug Discov 23:141−155. DOI:10.1038/s41573-023-00832-0 View in Article CrossRef Google Scholar
[14]	Grisoni, F., Moret, M., Lingwood, R., and Schneider, G. (2020). Bidirectional Molecule Generation with Recurrent Neural Networks. J Chem Inf Model 60:1175−1183. DOI:10.1021/acs.jcim.9b00943 View in Article CrossRef Google Scholar
[15]	Tang, Y., Moretti, R., and Meiler, J. (2024). Recent Advances in Automated Structure-Based De Novo Drug Design. J Chem Inf Model 64:1794−1805. DOI:10.1021/acs.jcim.4c00247 View in Article CrossRef Google Scholar
[16]	Jiang, Y., Zhang, G., You, J., et al. (2024). PocketFlow is a data-and-knowledge-driven structure-based molecular generative model. Nat Mach Intell 6:326−337. DOI:10.1038/s42256-024-00808-8 View in Article CrossRef Google Scholar
[17]	Schneuing, A., Harris, C., Du, Y., et al. (2024). Structure-based drug design with equivariant diffusion models. Nat Comput Sci 4:899−909. DOI:10.1038/s43588-024-00737-x View in Article CrossRef Google Scholar
[18]	Xu, M., Ran, T., and Chen, H. (2021). De Novo Molecule Design Through the Molecular Generative Model Conditioned by 3D Information of Protein Binding Sites. J Chem Inf Model 61:3240−3254. DOI:10.1021/acs.jcim.0c01494 View in Article CrossRef Google Scholar
[19]	Wu, K., Xia, Y., Deng, P., et al. (2024). TamGen: drug design with target-aware molecule generation through a chemical language model. Nat Commun 15:9360. DOI:10.1038/s41467-024-53632-4 View in Article CrossRef Google Scholar
[20]	Zhang, W., Zhang, K., and Huang, J. (2023). A Simple Way to Incorporate Target Structural Information in Molecular Generative Models. J Chem Inf Model 63:3719−3730. DOI:10.1021/acs.jcim.3c00293 View in Article CrossRef Google Scholar
[21]	Ünlü, A., Çevrim, E., Yiğit, M.G., et al. (2025). Target-specific de novo design of drug candidate molecules with graph-transformer-based generative adversarial networks. Nat Mach Intell 7:1524−1540. DOI:10.1038/s42256-025-01082-y View in Article CrossRef Google Scholar
[22]	Peng, X., Luo, S., Guan, J., et al. (2025). Pocket2Mol: Efficient Molecular Sampling Based on 3D Protein Pockets. DOI: 10.48550/arXiv.2205.07249 View in Article Google Scholar
[23]	Zhang, O., Zhang, J., Jin, J., et al. (2023). ResGen is a pocket-aware 3D molecular generation model based on parallel multiscale modelling. Nat Mach Intell 5:1020−1030. DOI:10.1038/s42256-023-00712-7 View in Article CrossRef Google Scholar
[24]	Hu, Q., Sun, C., He, H., et al. (2025). Target-aware 3D molecular generation based on guided equivariant diffusion. Nat Commun 16:7928. DOI:10.1038/s41467-025-63245-0 View in Article CrossRef Google Scholar
[25]	Huang, L., Xu, T., Yu, Y., et al. (2024). A dual diffusion model enables 3D molecule generation and lead optimization based on target pockets. Nat Commun 15:2657. DOI:10.1038/s41467-024-46569-1 View in Article CrossRef Google Scholar
[26]	Zhung, W., Kim, H., and Kim, W.Y. (2024). 3D molecular generative framework for interaction-guided drug design. Nat Commun 15:2688. DOI:10.1038/s41467-024-47011-2 View in Article CrossRef Google Scholar
[27]	Yu, J.-L., Zhou, C., Ning, X.-L., et al. (2025). Knowledge-guided diffusion model for 3D ligand-pharmacophore mapping. Nat Commun 16:2269. DOI:10.1038/s41467-025-57485-3 View in Article CrossRef Google Scholar
[28]	Ghorbani, M., Gendelev, L., Beroza, P., and Keiser, M.J. (2023). Autoregressive fragment-based diffusion for pocket-aware ligand design. DOI: 10.48550/arXiv.2401.05370 View in Article Google Scholar
[29]	Peng, J., Yu, J.-L., Yang, Z.-B., et al. (2025). Pharmacophore-oriented 3D molecular generation toward efficient feature-customized drug discovery. Nat Comput Sci 5:898–914. DOI: 10.1038/s43588-025-00850-5 View in Article Google Scholar
[30]	Yang, Y., Hsieh, C.-Y., Kang, Y., et al. (2023). Deep Generation Model Guided by the Docking Score for Active Molecular Design. J Chem Inf Model 63:2983−2991. DOI:10.1021/acs.jcim.3c00572 View in Article CrossRef Google Scholar
[31]	Dong, T., You, L., and Chen, C.Y.-C. (2025). Multi-objective drug design with a scaffold-aware variational autoencoder. Chem Sci 16:13352−13367. DOI:10.1039/D4SC08736D View in Article CrossRef Google Scholar
[32]	Xiong, G., Wu, Z., Yi, J., et al. (2021). ADMETlab 2.0: an integrated online platform for accurate and comprehensive predictions of ADMET properties. Nucleic Acids Res 49:W5-W14. DOI: 10.1093/nar/gkab255 View in Article Google Scholar
[33]	Kotsias, P.-C., Arús-Pous, J., Chen, H., et al. (2020). Direct steering of de novo molecular generation with descriptor conditional recurrent neural networks. Nat Mach Intell 2:254−265. DOI:10.1038/s42256-020-0174-5 View in Article CrossRef Google Scholar
[34]	Degen, J., Wegscheid-Gerlach, C., Zaliani, A., and Rarey, M. (2008). On the Art of Compiling and Using 'Drug-Like' Chemical Fragment Spaces. ChemMedChem 3:1503−1507. DOI:10.1002/cmdc.200800178 View in Article CrossRef Google Scholar
[35]	Pang, C., Qiao, J., Zeng, X., et al. (2024). Deep Generative Models in De Novo Drug Molecule Generation. J Chem Inf Model 64:2174−2194. DOI:10.1021/acs.jcim.3c01496 View in Article CrossRef Google Scholar
[36]	Schake, P., Bolz, S.N., Linnemann, K., and Schroeder, M. (2025). PLIP 2025: introducing protein–protein interactions to the protein–ligand interaction profiler. Nucleic Acids Res 53:W463−W465. DOI:10.1093/nar/gkaf361 View in Article CrossRef Google Scholar
[37]	Liu, Z., Su, M., Han, L., et al. (2017). Forging the Basis for Developing Protein–Ligand Interaction Scoring Functions. Accounts Chem Res 50:302−309. DOI:10.1021/acs.accounts.6b00491 View in Article CrossRef Google Scholar
[38]	Moret, M., Pachon Angona, I., Cotos, L., et al. (2023). Leveraging molecular structure and bioactivity with chemical language models for de novo drug design. Nat Commun 14:114. DOI:10.1038/s41467-022-35692-6 View in Article CrossRef Google Scholar
[39]	Bagal, V., Aggarwal, R., Vinod, P.K., and Priyakumar, U.D. (2022). MolGPT: Molecular Generation Using a Transformer-Decoder Model. J Chem Inf Model 62:2064−2076. DOI:10.1021/acs.jcim.1c00600 View in Article CrossRef Google Scholar
[40]	Grechishnikova, D. (2021). Transformer neural network for protein-specific de novo drug generation as a machine translation problem. Sci Rep 11:321. DOI:10.1038/s41598-020-79682-4 View in Article CrossRef Google Scholar
[41]	Chen, Y., Wang, Z., Wang, L., et al. (2023). Deep generative model for drug design from protein target sequence. J Cheminformatics 15:38. DOI:10.1186/s13321-023-00702-2 View in Article CrossRef Google Scholar
[42]	Born, J., and Manica, M. (2023). Regression Transformer enables concurrent sequence regression and generation for molecular language modelling. Nat Mach Intell 5:432−444. DOI:10.1038/s42256-023-00639-z View in Article CrossRef Google Scholar
[43]	Zhang, H., Song, H., Li, S., et al. (2023). A Survey of Controllable Text Generation Using Transformer-based Pre-trained Language Models. ACM Comput Surv 56:64:61–64:37. DOI: 10.1145/3617680. View in Article Google Scholar
[44]	Atz, K., Cotos, L., Isert, C., et al. (2024). Prospective de novo drug design with deep interactome learning. Nat Commun 15:3408. DOI:10.1038/s41467-024-47613-w View in Article CrossRef Google Scholar
[45]	Li, Y., Zhang, L., Wang, Y., et al. (2022). Generative deep learning enables the discovery of a potent and selective RIPK1 inhibitor. Nat Commun 13:6891. DOI:10.1038/s41467-022-34692-w View in Article CrossRef Google Scholar
[46]	Vaswani, A., Shazeer, N., Parmar, N., et al. (2023). Attention Is All You Need. DOI: 10.48550/arXiv.1706.03762. View in Article Google Scholar
[47]	Ross, J., Belgodere, B., Chenthamarakshan, V., et al. (2022). Large-scale chemical language representations capture molecular structure and properties. Nat Mach Intell 4:1256−1264. DOI:10.1038/s42256-022-00580-7 View in Article CrossRef Google Scholar
[48]	Su, M., Yang, Q., Du, Y., et al. (2019). Comparative Assessment of Scoring Functions: The CASF-2016 Update. J Chem Inf Model 59:895−913. DOI:10.1021/acs.jcim.8b00545 View in Article CrossRef Google Scholar
[49]	Yang, Y., Chen, G., Li, J., et al. (2024). Enabling target-aware molecule generation to follow multi objectives with Pareto MCTS. Commun Biol 7:1074. DOI:10.1038/s42003-024-06746-w View in Article CrossRef Google Scholar
[50]	Liu, H., Qin, Y., Niu, Z., et al. (2024). How Good are Current Pocket-Based 3D Generative Models?: The Benchmark Set and Evaluation of Protein Pocket-Based 3D Molecular Generative Models. J Chem Inf Model 64:9260-9275. DOI: 10.1021/acs.jcim.4c01598. View in Article Google Scholar
[51]	Jocys, Z., Grundy, J., and Farrahi, K. (2024). DrugPose: benchmarking 3D generative methods for early stage drug discovery. Digit Discov 3:1308-1318. DOI: https://doi.org/10.1039/d4dd00076e. View in Article Google Scholar
[52]	Baumgartner, C.K., Ebrahimi-Nik, H., Iracheta-Vellve, A., et al. (2023). The PTPN2/PTPN1 inhibitor ABBV-CLS-484 unleashes potent anti-tumour immunity. Nature 622:850−862. DOI:10.1038/s41586-023-06575-7 View in Article CrossRef Google Scholar
[53]	Gong, G.Q., Bilanges, B., Allsop, B., et al. (2023). A small-molecule PI3Kα activator for cardioprotection and neuroregeneration. Nature 618:159−168. DOI:10.1038/s41586-023-05972-2 View in Article CrossRef Google Scholar
[54]	Wang, Y., Zhang, J., Wu, X., et al. (2025). The Potential of PARP Inhibitors as Antitumor Drugs and the Perspective of Molecular Design. J Med Chem 68:18−48. DOI:10.1021/acs.jmedchem.4c02642 View in Article CrossRef Google Scholar
[55]	Baell, J.B., and Holloway, G.A. (2010). New Substructure Filters for Removal of Pan Assay Interference Compounds (PAINS) from Screening Libraries and for Their Exclusion in Bioassays. J Med Chem 53:2719−2740. DOI:10.1021/jm901137j View in Article CrossRef Google Scholar
[56]	Passaro, S., Corso, G., Wohlwend, J., et al. (2025). Boltz-2: Towards Accurate and Efficient Binding Affinity Prediction. DOI: 10.1101/2025.06.14.659707. View in Article Google Scholar
[57]	Yang, Z., Sun, W., Huang, Q., et al. (2026). Deep learning-assisted discovery of a potent and cell-active inhibitor of RNA N6-methyladenosine recognition protein YTHDC2. Nat Commun 17:46. DOI:10.1038/s41467-025-65542-0 View in Article CrossRef Google Scholar
[58]	Polykovskiy, D., Zhebrak, A., Sanchez-Lengeling, B., et al. (2020). Frontiers \| Molecular Sets (MOSES): A Benchmarking Platform for Molecular Generation Models. Front Pharmacol 11:565644. DOI: 10.3389/fphar.2020.565644. View in Article Google Scholar
[59]	Brown, N., Fiscato, M., Segler, M.H.S., and Vaucher, A.C. (2019). GuacaMol: Benchmarking Models for de Novo Molecular Design. J Chem Inf Model 59:1096−1108. DOI:10.1021/acs.jcim.8b00839 View in Article CrossRef Google Scholar
[60]	Wang, R., Fang, X., Lu, Y., et al. (2005). The PDBbind Database: Methodologies and Updates. J Med Chem 48:4111−4119. DOI:10.1021/jm048957q View in Article CrossRef Google Scholar
[61]	Liu, Z., Li, Y., Han, L., et al. (2014). PDB-wide collection of binding data: current status of the PDBbind database. Bioinformatics 31:405−412. DOI:10.1093/bioinformatics/btu626 View in Article CrossRef Google Scholar
[62]	Desaphy, J., Bret, G., Rognan, D., and Kellenberger, E. (2015). sc-PDB: a 3D-database of ligandable binding sites—10 years on. Nucleic Acids Res 43:D399−D404. DOI:10.1093/nar/gku928 View in Article CrossRef Google Scholar
[63]	Graber, D., Stockinger, P., Meyer, F., et al. (2025). Resolving data bias improves generalization in binding affinity prediction. Nat Mach Intell 7:1713−1725. DOI:10.1038/s42256-025-01124-5 View in Article CrossRef Google Scholar
[64]	Simonovsky, M., and Meyers, J. (2020). DeeplyTough: Learning Structural Comparison of Protein Binding Sites. J Chem Inf Model 60:2356−2366. DOI:10.1021/acs.jcim.9b00554 View in Article CrossRef Google Scholar
[65]	Liu, T., Naderi, M., Alvin, C., et al. (2017). Break Down in Order To Build Up: Decomposing Small Molecules for Fragment-Based Drug Design with eMolFrag. J Chem Inf Model 57:627−631. DOI:10.1021/acs.jcim.6b00596 View in Article CrossRef Google Scholar

Cited by

ABOUT THIS ARTICLE

Cite this article:

Qian R., Duan J., Liu X., et al. (2026). Structure–function conditioned molecular generation with PocketGPT enables prospective discovery of PARP1 inhibitors. The Innovation Drug Discovery 1:100028. https://doi.org/10.59717/j.xinn-drugdisc.2026.100028

Qian R., Duan J., Liu X., et al. (2026). Structure–function conditioned molecular generation with PocketGPT enables prospective discovery of PARP1 inhibitors. The Innovation Drug Discovery 1:100028. https://doi.org/10.59717/j.xinn-drugdisc.2026.100028

Welcome!

To request copyright permission to republish or share portions of our works, please visit Copyright Clearance Center's (CCC) Marketplace website at marketplace.copyright.com.

PDF

Figures(7) Tables(3)

Supplementary Information

download_supplementary
SI-Extended-TIDD-DOI-28

Article Metrics

Article views(286) PDF downloads(147)

Structure–function conditioned molecular generation with PocketGPT enables prospective discovery of PARP1 inhibitors

GRAPHICAL ABSTRACT

PUBLIC SUMMARY

ABSTRACT