Abstract
DNA-encoded libraries (DELs) facilitate high-throughput screening of trillions of molecules against protein targets through split-pool synthesis and DNA tagging. Despite their potential, only a few DEL-derived compounds have advanced to clinical trials or reached the market. A better understanding of the defining characteristics of target proteins, particularly those with binding pockets suitable for DEL screening, is critical to improving success rates. However, existing approaches remain limited in assessing pocket flexibility and functional similarity. Here, we present ErePOC, a pocket representation model based on contrastive learning with ESM-2 embeddings to address these challenges. ErePOC captures both structural and functional features of binding pockets, enabling identification of shared characteristics among DEL targets. By integrating analyses of low-dimensional physicochemical properties and high-dimensional ErePOC embeddings, we provide a comprehensive view of DEL target space. With 98% precision in downstream classification tasks, ErePOC demonstrates high performance in pocket representation, which is then applied to predict human proteins suitable for DEL screening, with enrichment uncovered across 18 functional categories. This work establishes a framework for enhancing DEL-based drug discovery through more effective target selection and pocket similarity analysis.
Similar content being viewed by others
Data availability
The datasets generated and analyzed in this study have been deposited in the Zenodo database under [https://doi.org/10.5281/zenodo.18033921]. The raw data supporting the findings of this study, including model outputs and processed datasets, are available from Zenodo under this accession code. Source data underlying all quantitative figures with manageable size are provided as individual Source data Excel files with this manuscript, including Figs. 1, 2, 5 and 7, and Supplementary Figs. S1–S4, S9, S12–S13, S15 and S17–S19. Due to the large scale of the datasets underlying the t-SNE visualizations (Fig. 4 and Supplementary Figs. S5–S11 and S14), individual data points are not provided as Source data files; however, the full processed inputs required to reproduce these figures are available on Zenodo. Raw data related to Figs. 1, 2, 7, S2, S4, S13, S17 and S19 are also provided on Zenodo. PyMOL session files (.pse) used for structural visualization in Fig. 6 and Supplementary Fig. S16 are available on Zenodo as well. The BioLiP2, AlphaFill, and AF2-predicted protein structure data used in this study are publicly available from the BioLiP, AlphaFill, and AlphaFold databases at [https://zhanggroup.org/BioLiP/], [https://alphafill.eu/], and [https://alphafold.ebi.ac.uk/download], respectively. Lists of target proteins for BioLiP2, AlphaFill, and AlphaFold-predicted human proteins, as well as PDB code mappings for DEL and FDA-AD targets, are also available on Zenodo. Source data are provided with this paper.
Code availability
All custom source code and algorithms generated in this study are publicly available at https://github.com/JingHuangLab/ErePOC, along with the processed data required to reproduce all results reported in this manuscript. There are no restrictions on access.
References
Ma, P. et al. Evolution of chemistry and selection technology for DNA-encoded library. Acta Pharm. Sin. B 14, 492–516 (2024).
Gironda-Martínez, A., Donckele, E. J., Samain, F. & Neri, D. DNA-encoded chemical libraries: a comprehensive review with successful stories and future challenges. ACS Pharmacol. Transl. Sci. 4, 1265–1279 (2021).
Hou, N. et al. Development of highly potent noncovalent inhibitors of SARS-CoV-2 3CLpro. ACS Cent. Sci. 9, 217–227 (2023).
Ding, Y. et al. Discovery of soluble epoxide hydrolase inhibitors through DNA-encoded library technology (ELT). Bioorg. Med. Chem. 41, 116216 (2021).
Cuozzo, J. W. et al. Novel autotaxin inhibitor for the treatment of idiopathic pulmonary fibrosis: a clinical candidate discovered using DNA-encoded chemistry. J. Med. Chem. 63, 7840–7856 (2020).
Harris, P. A. et al. Discovery of a first-in-class receptor interacting protein 1 (RIP1) kinase specific clinical candidate (GSK2982772) for the treatment of inflammatory diseases. J. Med. Chem. 60, 1247–1261 (2017).
Peterson, A. A. & Liu, D. R. Small-molecule discovery through DNA-encoded libraries. Nat. Rev. Drug Discov. 22, 699–722 (2023).
Wichert, M., Guasch, L. & Franzini, R. M. Challenges and Prospects of DNA-Encoded Library Data Interpretation. Chem. Rev. 124, 12551–12572 (2024).
Lin, Z. et al. Evolutionary-scale prediction of atomic-level protein structure with a language model. Science 379, 1123–1130 (2023).
Ouyang-Zhang, J., Diaz, D., Klivans, A. & Krähenbühl, P. Predicting a protein’s stability under a million mutations. Advances in Neural Information Processing Systems 36, 76229–76247 (NeurIPS, 2023).
Kulmanov, M. et al. Protein function prediction as approximate semantic entailment. Nat. Mach. Intell. 6, 220–228 (2024).
Wang, D. et al. S-PLM: structure-aware protein language model via contrastive learning between sequence and structure. Adv. Sci. 12, e2404212 (2025).
He, Y. et al. Protein language models-assisted optimization of a uracil-N-glycosylase variant enables programmable T-to-G and T-to-C base editing. Mol. Cell 84, 1257–1270 e1256 (2024).
Gainza, P. et al. Deciphering interaction fingerprints from protein molecular surfaces using geometric deep learning. Nat. Methods 17, 184–192 (2020).
Zhou, G. et al. Uni-Mol: a universal 3D molecular representation learning framework. The Eleventh International Conference on Learning Representations (OpenReview, 2023).
Li, S. et al. PocketAnchor: learning structure-based pocket representations for protein-ligand interaction prediction. Cell Syst. 14, 692–705. e696 (2023).
Luo, Z., Wu, W., Sun, Q. & Wang, J. Accurate and transferable drug–target interaction prediction with DrugLAMP. Bioinformatics 40, btae693 (2024).
Zhao, L., Wang, H. & Shi, S. PocketDTA: an advanced multimodal architecture for enhanced prediction of drug−target affinity from 3D structural data of target binding pockets. Bioinformatics 40, btae594 (2024).
Chen, T., Kornblith, S., Norouzi, M. & Hinton, G. A simple framework for contrastive learning of visual representations. International Conference on Machine Learning, 1597–1607 (PMLR, 2020).
Kahraman, A., Morris, R. J., Laskowski, R. A. & Thornton, J. M. Shape variation in protein binding pockets and their ligands. J. Mol. Biol. 368, 283–301 (2007).
Skolnick, J. & Gao, M. Interplay of physics and evolution in the likely origin of protein biochemical function. Proc. Natl. Acad. Sci. USA 110, 9344–9349 (2013).
Dhenni, R. et al. Macrophages direct location-dependent recall of B cell memory to vaccination. Cell 188, 3477–3496 e3422 (2025).
Zhang, C., Zhang, X., Freddolino, P. L. & Zhang, Y. BioLiP2: an updated structure database for biologically relevant ligand–protein interactions. Nucleic Acids Res. 52, D404–D412 (2024).
Hekkelman, M. L., de Vries, I., Joosten, R. P. & Perrakis, A. AlphaFill: enriching AlphaFold models with ligands and cofactors. Nat. Methods 20, 205–213 (2023).
Bickerton, G. R., Paolini, G. V., Besnard, J., Muresan, S. & Hopkins, A. L. Quantifying the chemical beauty of drugs. Nat. Chem. 4, 90–98 (2012).
Leeson, P. D. & Springthorpe, B. The influence of drug-like concepts on decision-making in medicinal chemistry. Nat. Rev. Drug Discov. 6, 881–890 (2007).
Le Guilloux, V., Schmidtke, P. & Tuffery, P. Fpocket: an open source platform for ligand pocket detection. BMC Bioinforma. 10, 1–11 (2009).
Jubb, H. C. et al. Arpeggio: a web server for calculating and visualising interatomic interactions in protein structures. J. Mol. Biol. 429, 365–371 (2017).
Vilar, S., Cozza, G. & Moro, S. Medicinal chemistry and the molecular operating environment (MOE): application of QSAR and molecular docking to drug discovery. Curr. Top. Med. Chem. 8, 1555–1572 (2008).
Van der Maaten, L. & Hinton, G. Visualizing data using t-SNE. J. Mach. Learn. Res. 9, 2579–2605 (2008).
Varadi, M. et al. AlphaFold Protein Structure Database in 2024: providing structure coverage for over 214 million protein sequences. Nucleic Acids Res. 52, D368–D375 (2024).
Skolnick, J. & Zhou, H. Implications of the essential role of small molecule ligand binding pockets in protein-protein interactions. J. Phys. Chem. B 126, 6853–6867 (2022).
Zhang, Y. & Skolnick, J. TM-align: a protein structure alignment algorithm based on the TM-score. Nucleic acids Res. 33, 2302–2309 (2005).
Hu, J. & Zhang, Y. PPS-align: an user-friendly pocket structural alignment server. https://zhanglab.dcmb.med.umich.edu/PPS-align/.
Xie, L. & Bourne, P. E. Detecting evolutionary relationships across existing fold space, using sequence order-independent profile-profile alignments. Proc. Natl. Acad. Sci. USA 105, 5441–5446 (2008).
Iqbal, S. et al. Evaluation of DNA encoded library and machine learning model combinations for hit discovery. npj Drug Discov. 2, 5 (2025).
Lyu, J. et al. Ultra-large library docking for discovering new chemotypes. Nature 566, 224–229 (2019).
Liu, F. et al. The impact of library size and scale of testing on virtual screening. Nat. Chem. Biol. 21, 1039–1045 (2025).
Pulous, F. E. et al. MAT2A inhibition combats metabolic and transcriptional reprogramming in cancer. Drug Discov. Today 29, 104189 (2024).
Zhang, W. & Huang, J. EViS: an enhanced virtual screening approach based on pocket–ligand similarity. J. Chem. Inf. Model. 62, 498–510 (2022).
Zhan, R., Zhang, W. & Huang, J. Scrutinization on docking against individually generated target pockets for each ligand. Preprint at bioRxiv https://doi.org/10.1101/2025.01.01.630989 (2025).
Nishimura, T., Tsuboyama, K., Nakagaki, Y., Yamamoto, E. & Mizushima, N. Rational engineering of lipid-binding probes via high-throughput protein-lipid interaction screening. Preprint at bioRxiv https://doi.org/10.1101/2025.01.05.627504 (2025).
Daina, A. & Zoete, V. Testing the predictive power of reverse screening to infer drug targets, with the help of machine learning. Commun. Chem. 7, 105 (2024).
Zhang, W., Zhang, K. & Huang, J. A simple way to incorporate target structural information in molecular generative models. J. Chem. Inf. Model. 63, 3719–3730 (2023).
Watson, J. L. et al. De novo design of protein structure and function with RFdiffusion. Nature 620, 1089–1100 (2023).
Abramson, J. et al. Addendum: accurate structure prediction of biomolecular interactions with AlphaFold 3. Nature 636, E4 (2024).
Chen, G. et al. Rapid clearance of achiral small-molecule drugs using de novo-designed proteins and their cyclic and mirror-image variants. Nat. Biomed. Eng. 9, 1775–1787 (2025).
Polizzi, N. F. & DeGrado, W. F. A defined structural unit enables de novo design of small-molecule-binding proteins. Science 369, 1227–1233 (2020).
Bento, A. P. et al. An open source chemical structure curation pipeline using RDKit. J. Cheminform. 12, 51 (2020).
Subramanian, A. et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc. Natl. Acad. Sci. USA 102, 15545–15550 (2005).
O’Boyle, N. M. et al. Open Babel: an open chemical toolbox. J. Cheminform. 3, 33 (2011).
Trott, O. & Olson, A. J. AutoDock Vina: improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading. J. Comput. Chem. 31, 455–461 (2010).
Acknowledgements
This work was supported by the “Pioneer” and “Leading Goose” R&D Program of Zhejiang, grant numbers 2023C03109 (J.H.) and 2024SSYS0036 (J.H.); the National Natural Science Foundation of China, grant 32171247 (J.H.), T2596084 (J.H.), and 32501101 (W.Z.); the Zhejiang Provincial Natural Science Foundation, grant LQ23F020011 (W.Z.); the State Key Laboratory of Gene Expression; and the Westlake Education Foundation. The authors thank the Westlake University Supercomputer Center for computational resources and related assistance.
Author information
Authors and Affiliations
Contributions
J.H. and Q.H. conceived the study. W.Z., Y.W., R.Z., and R.Q. designed the experiments. All authors analyzed the results. J.H. and W.Z. wrote the manuscript.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Peer review
Peer review information
Nature Communications thanks Holly Soutter, Noel O’Boyle, who co-reviewed with Melissa F. Adasme, 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.
Supplementary information
Source data
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, 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 changes were made. 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/4.0/.
About this article
Cite this article
Zhang, W., Wang, Y., Zhan, R. et al. Deciphering DEL pocket patterns through contrastive learning. Nat Commun (2026). https://doi.org/10.1038/s41467-026-69663-y
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41467-026-69663-y
