LeadFlow learns activity-aligned molecular generation from chemical series for structure-based molecular optimization
Abstract
Structure-based molecular optimization requires supervision that captures scaffold modifications associated with experimental activity. Here we introduce LeadFlow, combining maximum common substructure supervision from assay-consistent chemical series, Bayesian Flow Networks for joint generation of three-dimensional coordinates and molecular graphs, and experimental activity-based direct preference optimization (DPO). On MolGenBench, LeadFlow recovered reference active scaffolds in approximately 400 of 600 series and exact active molecules in approximately 120 series. Controlled scaffold-definition ablations supported broader active-chemotype recovery and fused-ring elaboration. Activity-based DPO increased activity-weighted recovery relative to pretraining and docking-score-based preferences. In a prospective ClpP campaign, generation followed by computational filtering and medicinal chemistry selection yielded five synthesized compounds with lower mean EC₅₀ values than the starting compound A1. DC-ClpP-03 reached 38.7 nM, an approximately 156-fold potency improvement. These results support chemical-series supervision and experimental preferences for scaffold-conditioned molecular optimization.
References
1. Krishnan, A., et al. A generative deep learning approach to de novo antibiotic design. Cell 188, 5962–5979.e5922 (2025).
2. Zhang, K., et al. Artificial intelligence in drug development. Nature Medicine 31, 45–59 (2025).
3. Xu, Z., et al. A generative AI-discovered TNIK inhibitor for idiopathic pulmonary fibrosis: a randomized phase 2a trial. Nature Medicine 31, 2602–2610 (2025).
4. Watson, J.L., et al. De novo design of protein structure and function with RFdiffusion. Nature 620, 1089–1100 (2023).
5. Abramson, J., et al. Accurate structure prediction of biomolecular interactions with AlphaFold 3. Nature 630, 493–500 (2024).
6. Liu, H., et al. How Good are Current Pocket-Based 3D Generative Models?: The Benchmark Set and Evaluation of Protein Pocket-Based 3D Molecular Generative Models. Journal of Chemical Information and Modeling 64, 9260–9275 (2024).
7. Cao, D., et al. Evaluating generative models for structure-based drug discovery from de novo design to lead optimization. Nature Communications (2026).
8. Gao, W. & Coley, C.W. The synthesizability of molecules proposed by generative models. Journal of chemical information and modeling 60, 5714–5723 (2020).
9. Yang, B., Xiang, C., Li, T., Xu, Y. & Li, J. Structure-Based Generation of 3D Small- Molecule Drugs: Are We There Yet? Journal of Medicinal Chemistry 68, 22756–22768 (2025).
10.Zhang, O., et al. Deep Lead Optimization: Leveraging Generative AI for Structural Modification. Journal of the American Chemical Society 146, 31357–31370 (2024).
11.Langevin, M., Minoux, H., Levesque, M. & Bianciotto, M. Scaffold-Constrained Molecular Generation. Journal of Chemical Information and Modeling 60, 5637–5646 (2020).
12.Liu, X., Ye, K., van Vlijmen, H.W.T., Ijzerman, A.P. & van Westen, G.J.P. DrugEx v3: scaffold-constrained drug design with graph transformer-based reinforcement learning. Journal of Cheminformatics 15, 24 (2023).
13.Adams, K., Abeywardane, K., Fromer, J.C. & Coley, C.W. ShEPhERD: Diffusing shape, electrostatics, and pharmacophores for bioisosteric drug design. in ICLR (2025).
14.Zhu, H., Zhou, R., Cao, D., Tang, J. & Li, M. A pharmacophore-guided deep learning approach for bioactive molecular generation. Nature Communications 14, 6234 (2023).
15.Xie, W., et al. Accelerating discovery of bioactive ligands with pharmacophore-informed generative models. Nature Communications 16, 2391 (2025).
16.Chen, S., et al. Deep lead optimization enveloped in protein pocket and its application in designing potent and selective ligands targeting LTK protein. Nature Machine Intelligence 7, 448–458 (2025).
17.Xie, J., Chen, S., Lei, J. & Yang, Y. DiffDec: structure-aware scaffold decoration with an end-to-end diffusion model. Journal of Chemical Information and Modeling 64, 2554–2564 (2024).
18.Schneuing, A., et al. Structure-based drug design with equivariant diffusion models. Nature Computational Science 4, 899–909 (2024).
19.Lin, H., et al. CBGBench: Fill in the blank of protein-molecule complex binding graph. arXiv preprint arXiv:2406.10840 (2024).
20.Francoeur, P.G., et al. Three-Dimensional Convolutional Neural Networks and a Cross- Docked Data Set for Structure-Based Drug Design. Journal of Chemical Information and Modeling 60, 4200–4215 (2020).
21.Lewell, X.Q., Judd, D.B., Watson, S.P. & Hann, M.M. Recap retrosynthetic combinatorial analysis procedure: a powerful new technique for identifying privileged molecular fragments with
useful applications in combinatorial chemistry. Journal of chemical information and computer sciences 38, 511–522 (1998).
22.Arús-Pous, J., et al. SMILES-based deep generative scaffold decorator for de-novo drug design. Journal of cheminformatics 12, 38 (2020).
23.Degen, J., Wegscheid-Gerlach, C., Zaliani, A. & Rarey, M. On the art of compiling and using'drug-like'chemical fragment spaces. ChemMedChem 3, 1503 (2008).
24.Cremer, J., et al. FLOWR: flow matching for structure-aware de novo, interaction- and fragment-based ligand generation. Nature Computational Science 6, 565–574 (2026).
25.Peng, X., et al. Unified modeling of 3D molecular generation via atomic interactions with PocketXMol. Cell 189, 1904–1922.e1928 (2026).
26.Kong, X., et al. Programming Biomolecular Interactions with All-Atom Generative Model. bioRxiv, 2026.2003.2012.711044 (2026).
27.Qiao, A., et al. A 3D pocket-aware lead optimization model with knowledge guidance and its application for discovery of new glutaminyl cyclase inhibitors. Briefings in Bioinformatics 26, bbaf345 (2025).
28.Gu, S., et al. Aligning target-aware molecule diffusion models with exact energy optimization. Advances in Neural Information Processing Systems 37, 44040–44063 (2024).
29.Qiu, K., et al. Empower structure-based molecule optimization with gradient guided bayesian flow networks. arXiv preprint arXiv:2411.13280 (2024).
30.Black, K., Janner, M., Du, Y., Kostrikov, I. & Levine, S. Training diffusion models with reinforcement learning. in International Conference on Learning Representations, Vol. 2024 4965–4987 (2024).
31.Stacey, J., Canault, B., Pickett, S.D. & Gillet, V.J. Visualising lead optimisation series using reduced graphs. Journal of Cheminformatics 17, 60 (2025).
32.Takeuchi, K., Kunimoto, R. & Bajorath, J. Systematic mapping of R-group space enables the generation of an R-group replacement system for medicinal chemistry. European Journal of Medicinal Chemistry 225, 113771 (2021).
33.Hoffer, L., Muller, C., Roche, P. & Morelli, X. Chemistry-driven Hit-to-lead Optimization Guided by Structure-based Approaches. Molecular Informatics 37, 1800059 (2018).
34.Yu, J., et al. Atomic-level protein–ligand recognition with PBCNet2.0 for probe discovery. Nature Chemical Biology 22, 1299–1310 (2026).
35.Yu, J., et al. Computing the relative binding affinity of ligands based on a pairwise binding 1comparison network. Nature computational science 3, 860–872 (2023).
36.Graves, A., Srivastava, R.K., Atkinson, T. & Gomez, F. Bayesian flow networks. arXiv preprint arXiv:2308.07037 (2023). 1215
37.Wallace, B., et al. Diffusion model alignment using direct preference optimization. in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition 8228– 8238 (2024).
38.Rafailov, R., et al. Direct preference optimization: Your language model is secretly a reward model. Advances in neural information processing systems 36, 53728–53741 (2023).
39.Song, Y., et al. Unified generative modeling of 3d molecules with bayesian flow networks. in The Twelfth International Conference on Learning Representations (2024).
40.Qiu, K., et al. Piloting structure-based drug design via modality-specific optimal schedule. arXiv preprint arXiv:2505.07286 (2025).
41.Qu, Y., et al. MolCRAFT: structure-based drug design in continuous parameter space. arXiv preprint arXiv:2404.12141 (2024).
42.Atkinson, T., et al. Protein sequence modelling with Bayesian flow networks. Nature Communications 16, 3197 (2025).
43.Ruple, L., Torresi, L., Schopmans, H. & Friederich, P. Symmetry-aware Bayesian flow networks for crystal generation. npj Computational Materials 12, 182 (2026).
44.Zeng, C., et al. Structure-free generation of transition state conformations with Bayesian flow networks. ChemRxiv 2026(2026).
45.Jiao, R., et al. Mof-bfn: Metal-organic frameworks structure prediction via bayesian flow networks. Advances in Neural Information Processing Systems 38, 161989–162015 (2026).
46.Kadan, A., Ryczko, K., Lloyd, E., Roitberg, A. & Yamazaki, T. Guided multi-objective generative AI to enhance structure-based drug design. Chemical Science 16, 13196–13210 (2025).
47.Schindler, C.E., et al. Large-scale assessment of binding free energy calculations in active drug discovery projects. Journal of Chemical Information and Modeling 60, 5457–5474 (2020).
48.Wang, L., et al. Accurate and Reliable Prediction of Relative Ligand Binding Potency in Prospective Drug Discovery by Way of a Modern Free-Energy Calculation Protocol and Force Field. Journal of the American Chemical Society 137, 2695–2703 (2015).
49.Chen, Z., Peng, B., Parthasarathy, S. & Ning, X. Shape-conditioned 3D molecule generation via equivariant diffusion models. arXiv preprint arXiv:2308.11890 (2023).
50.Harris, C., et al. Posecheck: Generative models for 3d structure-based drug design produce unrealistic poses. in NeurIPS 2023 Generative AI and Biology (GenBio) Workshop (2023).
51.Buttenschoen, M., Morris, G.M. & Deane, C.M. PoseBusters: AI-based docking methods fail to generate physically valid poses or generalise to novel sequences. Chemical Science 15, 3130–3139 (2024).
52.Trott, O. & Olson, A.J. AutoDock Vina: improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading. Journal of computational chemistry 31, 455–461 (2010).
53.Bouysset, C. & Fiorucci, S. ProLIF: a library to encode molecular interactions as fingerprints. Journal of cheminformatics 13, 72 (2021).
54.Bothe, U., et al. Discovery of IRAK4 Inhibitors BAY1834845 (Zabedosertib) and BAY1830839. Journal of Medicinal Chemistry 67, 1225–1242 (2024).
55.Ngo, V.N., et al. Oncogenically active MYD88 mutations in human lymphoma. Nature 470, 115–119 (2011).
56.Bai, Y.-R., et al. The recent advance of Interleukin-1 receptor associated kinase 4 inhibitors for the treatment of inflammation and related diseases. European Journal of Medicinal Chemistry 258, 115606 (2023).
57.Chen, Y., et al. Bicyclic pyrimidine compounds as potent IRAK4 inhibitors. Bioorganic & Medicinal Chemistry Letters 73, 128900 (2022).
58.Haynes, C.M., Petrova, K., Benedetti, C., Yang, Y. & Ron, D. ClpP Mediates Activation of a Mitochondrial Unfolded Protein Response in C. elegans. Developmental Cell 13, 467–480 (2007).
59.Annesley, S.J. & Fisher, P.R. Mitochondria in Health and Disease. Cells 8, 680 (2019).
60.Topf, U., Wrobel, L. & Chacinska, A. Chatty Mitochondria: Keeping Balance in Cellular Protein Homeostasis. Trends in Cell Biology 26, 577–586 (2016).
61.Vyas, S., Zaganjor, E. & Haigis, M.C. Mitochondria and Cancer. Cell 166, 555–566 (2016).
62.Ishizawa, J., et al. Mitochondrial ClpP-Mediated Proteolysis Induces Selective Cancer Cell Lethality. Cancer Cell 35, 721–737.e729 (2019).
63.Suomalainen, A. & Battersby, B.J. Mitochondrial diseases: the contribution of organelle stress responses to pathology. Nature Reviews Molecular Cell Biology 19, 77–92 (2018).
64.Cole, A., et al. Inhibition of the Mitochondrial Protease ClpP as a Therapeutic Strategy for Human Acute Myeloid Leukemia. Cancer Cell 27, 864–876 (2015).
65.Uhlen, M., et al. A pathology atlas of the human cancer transcriptome. Science 357, eaan2507 (2017).
66.Mirali, S. & Schimmer, A.D. The role of mitochondrial proteases in leukemic cells and leukemic stem cells. Stem Cells Translational Medicine 9, 1481–1487 (2020).
67.Chen, S., et al. Design, Synthesis, and Structure-Activity Relationships of Novel Piperidine-Fused Imidazolone ClpP Activators as Potential Anti-Cancer Agents. Molecules 31, 2916 (2026).
68.Kumar, R., et al. Structure and Energetics of Ligand–Fluorine Interactions with Galectin‐3 Backbone and Side‐Chain Amides: Insight into Solvation Effects and Multipolar Interactions. ChemMedChem 14, 1528–1536 (2019).
69.Mabanglo, M.F., et al. Potent ClpP agonists with anticancer properties bind with improved structural complementarity and alter the mitochondrial N-terminome. Structure 31, 185–200.e110 (2023).
70.Zdrazil, B., et al. The ChEMBL Database in 2023: a drug discovery platform spanning multiple bioactivity data types and time periods. Nucleic acids research 52, D1180–D1192 (2024).
71.Ouyang, L., et al. Training language models to follow instructions with human feedback. Advances in neural information processing systems 35, 27730–27744 (2022).
72.Bradley, R.A. & Terry, M.E. Rank analysis of incomplete block designs: I. the method of paired comparisons. Biometrika 39, 324–345 (1952).
73.Zhou, X., et al. Antigen-specific antibody design via direct energy-based preference optimization. Advances in Neural Information Processing Systems 37, 120861–120891 (2024).
74.Lee, S., Jo, M., Ok, J. & Kim, D. Enhancing Ligand Validity and Affinity in Structure-Based Drug Design with Multi-Reward Optimization. in Forty-second International Conference on Machine Learning (2025).
75.Noutahi, E., et al. datamol-io/medchem: Molecular filtering for drug discovery. (Zenodo, 2025).
76.Baell, J.B. & Holloway, G.A. New Substructure Filters for Removal of Pan Assay Interference Compounds (PAINS) from Screening Libraries and for Their Exclusion in Bioassays. Journal of Medicinal Chemistry 53, 2719–2740 (2010).
77.Pearce, B.C., Sofia, M.J., Good, A.C., Drexler, D.M. & Stock, D.A. An Empirical Process for the Design of High-Throughput Screening Deck Filters. Journal of Chemical Information and Modeling 46, 1060–1068 (2006).
Metrics
DOI:
Submission ID:
Downloads
Posted
How to Cite
Download Citation
Declaration of Competing Interests
The authors declare no competing interests to disclose.
Copyright
The copyright holder for this preprint is the author/funder.
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.