Structure-based drug design with equivariant diffusion models

Abstract

Structure-based drug design (SBDD) aims to design small-molecule ligands that bind with high affinity and specificity to pre-determined protein targets. Generative SBDD methods leverage structural data of drugs with their protein targets to propose new drug candidates. However, most existing methods focus exclusively on bottom-up de novo design of compounds or tackle other drug development challenges with task-specific models. The latter requires curation of suitable datasets, careful engineering of the models and retraining from scratch for each task. Here we show how a single pretrained diffusion model can be applied to a broader range of problems, such as off-the-shelf property optimization, explicit negative design and partial molecular design with inpainting. We formulate SBDD as a three-dimensional conditional generation problem and present DiffSBDD, an SE(3)-equivariant diffusion model that generates novel ligands conditioned on protein pockets. Furthermore, we show how additional constraints can be used to improve the generated drug candidates according to a variety of computational metrics.

References 78

1. 1.

   Anderson AC. The process of structure-based drug design. Chem. Biol. 2003;10:787-797.

   Not in our library10.1016/j.chembiol.2003.09.002

2. 2.

   Lyne PD. Structure-based virtual screening: an overview. Drug Discov. Today. 2002;7:1047-1055.

   Not in our library10.1016/s1359-6446(02)02483-2

3. 3.

   Shoichet BK. Virtual screening of chemical libraries. Nature. 2004;432:862-865.

   Not in our library10.1038/nature03197

4. 4.

   Irwin JJ, Shoichet BK. ZINC—a free database of commercially available compounds for virtual screening. J. Chem. Inf. Model. 2005;45:177-182.

   Not in our library10.1021/ci049714+

5. 5.
   Molecular Docking and Structure-Based Drug Design Strategies · 2015

   No notice
6. 6.

   Bronstein, M. M., Bruna, J., Cohen, T. & Veličković, P. Geometric deep learning: grids, groups, graphs, geodesics, and gauges. Preprint at https://arxiv.org/abs/2104.13478 (2021).

   Not in our library

7. 7.

   Atz K, Grisoni F, Schneider G. Geometric deep learning on molecular representations. Nat. Mach. Intell. 2021;3:1023-1032.

   Not in our library10.1038/s42256-021-00418-8

8. 8.

   Khakzad H, et al. A new age in protein design empowered by deep learning. Cell Syst. 2023;14:925-939.

   Not in our library10.1016/j.cels.2023.10.006

9. 9.

   Gaudelet, T. et al. Utilizing graph machine learning within drug discovery and development. Brief. Bioinform.10.1093/bib/bbab159 (2021).

   Not in our library10.1093/bib/bbab159

10. 10.

    Lu, W. et al. TANKBind: trigonometry-aware neural networks for drug-protein binding structure prediction. In Advances in Neural Information Processing Systems 7236–7249 (Curran Associates, 2022).

    Not in our library

11. 11.

    Stärk, H., Ganea, O., Pattanaik, L., Barzilay, R. & Jaakkola, T. EquiBind: geometric deep learning for drug binding structure prediction. In International Conference on Machine Learning 20503–20521 (PMLR, 2022).

    Not in our library

12. 12.

    Corso, G., Jing, B., Barzilay, R. & Jaakkola, T. DiffDock: diffusion steps, twists, and turns for molecular docking. In International Conference on Learning Representations (OpenReview.net, 2023); https://openreview.net/forum?id=kKF8_K-mBbS

    Not in our library

13. 13.
    Generating 3D molecules conditional on receptor binding sites with deep generative models · 2022

    No notice
14. 14.

    Li Y, Pei J, Lai L. Structure-based de novo drug design using 3D deep generative models. Chem. Sci. 2021;12:13664-13675.

    Not in our library10.1039/d1sc04444c

15. 15.

    Peng, X. et al. Pocket2mol: efficient molecular sampling based on 3D protein pockets. In Proc. 39th International Conference on Machine Learning 17644–17655 (PMLR, 2022).

    Not in our library

16. 16.

    Drotár, P., Jamasb, A. R., Day, B., Cangea, C. & Liò, P. Structure-aware generation of drug-like molecules. Preprint at https://arxiv.org/abs/2111.04107 (2021).

    Not in our library

17. 17.

    Liu, M., Luo, Y., Uchino, K., Maruhashi, K. & Ji, S. Generating 3D molecules for target protein binding. In Proc. 39th International Conference on Machine Learning 13912–13924 (PMLR, 2022).

    Not in our library

18. 18.

    Guan, J. et al. 3D equivariant diffusion for target-aware molecule generation and affinity prediction. In Eleventh International Conference on Learning Representations (OpenReview.net, 2023); https://openreview.net/forum?id=kJqXEPXMsE0

    Not in our library

19. 19.

    Lin, H. et al. DiffBP: generative diffusion of 3D molecules for target protein binding. Preprint at https://arxiv.org/abs/2211.11214 (2022).

    Not in our library10.1039/d4sc05894a

20. 20.

    Guan, J. et al. DecompDiff: diffusion models with decomposed priors for structure-based drug design. In Proc. 40th International Conference on Machine Learning 11827–11846 (PMLR, 2023).

    Not in our library

21. 21.

    Xu, M., Powers, A. S., Dror, R. O., Ermon, S. & Leskovec, J. Geometric latent diffusion models for 3D molecule generation. In International Conference on Machine Learning 38592–38610 (PMLR, 2023).

    Not in our library

22. 22.

    Weiss, T. et al. Guided diffusion for inverse molecular design. Nat. Comput. Sci.3, 873–882 (2023).

    Not in our library10.1038/s43588-023-00532-0

23. 23.

    Ho J, Jain A, Abbeel P. Denoising diffusion probabilistic models. Adv. Neural Inf. Process. Syst. 2020;33:6840-6851.

    Not in our library

24. 24.

    Hoogeboom, E., Satorras, V. G., Vignac, C. & Welling, M. Equivariant diffusion for molecule generation in 3D. In International Conference on Machine Learning 8867–8887 (PMLR, 2022).

    Not in our library

25. 25.

    Zhang, O. et al. ResGen is a pocket-aware 3D molecular generation model based on parallel multiscale modelling. Nat. Mach. Intell.5, 1020–1030 (2023).

    Not in our library

26. 26.

    Jiang Y, et al. Pocketflow is a data-and-knowledge-driven structure-based molecular generative model. Nat. Mach. Intell. 2024;6:326-337.

    Not in our library10.1038/s42256-024-00808-8

27. 27.

    Zhung W, Kim H, Kim WY. 3D molecular generative framework for interaction-guided drug design. Nat. Commun. 2024;15:2688.

    Not in our library10.1038/s41467-024-47011-2

28. 28.

    Hu L, Benson ML, Smith RD, Lerner MG, Carlson HA. Binding MOAD (Mother Of All Databases). Proteins. 2005;60:333-340.

    Not in our library10.1002/prot.20512

29. 29.

    Chen P, et al. Structural basis for recognition of frizzled proteins by Clostridium difficile toxin B. Science. 2018;360:664-669.

    Not in our library10.1126/science.aar1999

30. 30.

    Ding L-C, et al. FZD2 inhibits the cell growth and migration of salivary adenoid cystic carcinomas. Oncol. Rep. 2016;35:1006-1012.

    Not in our library10.3892/or.2015.3811

31. 31.

    Ritchie TJ, Macdonald SJ. The impact of aromatic ring count on compound developability—are too many aromatic rings a liability in drug design? Drug Discov. Today. 2009;14:1011-1020.

    Not in our library10.1016/j.drudis.2009.07.014

32. 32.

    Song, Y. et al. Score-based generative modeling through stochastic differential equations. In International Conference on Learning Representations (OpenReview.net, 2021); https://openreview.net/forum?id=PxTIG12RRHS

    Not in our library

33. 33.

    Lugmayr, A. et al. Repaint: inpainting using denoising diffusion probabilistic models. In Proc. IEEE/CVF Conference on Computer Vision and Pattern Recognition 11461–11471 (IEEE, 2022).

    Not in our library

34. 34.
    Scaffolding protein functional sites using deep learning · 2022

    No notice
35. 35.

    Ho J, et al. Video diffusion models. Adv. Neural Inf. Process. Syst. 2022;35:8633-8646.

    Not in our library

36. 36.

    Igashov, I. et al. Equivariant 3D-conditional diffusion model for molecular linker design. Nat. Mach. Intell.6, 417–427 (2024).

    Not in our library

37. 37.

    Kingma D, Salimans T, Poole B, Ho J. Variational diffusion models. Adv. Neural Inf. Process. Syst. 2021;34:21696-21707.

    Not in our library

38. 38.

    Lepola U, Wade A, Andersen HF. Do equivalent doses of escitalopram and citalopram have similar efficacy? A pooled analysis of two positive placebo-controlled studies in major depressive disorder. Int. Clin. Psychopharmacol. 2004;19:149-155.

    Not in our library10.1097/00004850-200405000-00005

39. 39.

    Satorras, V. G., Hoogeboom, E. & Welling, M. E(n) equivariant graph neural networks. In International Conference on Machine Learning 9323–9332 (PMLR, 2021).

    Not in our library

40. 40.

    Köhler, J., Klein, L. & Noé, F. Equivariant flows: exact likelihood generative learning for symmetric densities. In International Conference on Machine Learning 5361–5370 (PMLR, 2020).

    Not in our library

41. 41.

    Serre, J.-P. Linear Representations of Finite Groups Vol. 42 (Springer, 1977).

    Not in our library

42. 42.

    Trippe, B. et al. Diffusion probabilistic modeling of protein backbones in 3D for the motif-scaffolding problem. In Eleventh International Conference on Learning Representations (OpenReview.net, 2023); https://openreview.net/forum?id=6TxBxqNME1Y

    Not in our library

43. 43.

    Nichol, A. Q. & Dhariwal, P. Improved denoising diffusion probabilistic models. In International Conference on Machine Learning 8162–8171 (PMLR, 2021).

    Not in our library

44. 44.

    Luo S, Guan J, Ma J, Peng J. A 3D generative model for structure-based drug design. Adv. Neural Inf. Process. Syst. 2021;34:6229-6239.

    Not in our library

45. 45.
    Open Babel: An open chemical toolbox · 2011

    No notice
46. 46.

    Bemis GW, Murcko MA. The properties of known drugs. 1. Molecular frameworks. J. Med. Chem. 1996;39:2887-2893.

    Not in our library10.1021/jm9602928

47. 47.

    Böhm H-J, Flohr A, Stahl M. Scaffold hopping. Drug Discov. Today Technol. 2004;1:217-224.

    Not in our library10.1016/j.ddtec.2004.10.009

48. 48.

    Kim KS, et al. Synthesis and sar of pyrrolotriazine-4-one based Eg5 inhibitors. Bioorg. Med. Chem. Lett. 2006;16:3937-3942.

    Not in our library10.1016/j.bmcl.2006.05.037

49. 49.

    Barone M, et al. Designed nanomolar small-molecule inhibitors of Ena/VASP EVH1 interaction impair invasion and extravasation of breast cancer cells. Proc. Natl Acad. Sci. USA. 2020;117:29684-29690.

    Not in our library10.1073/pnas.2007213117

50. 50.
    Application of Fragment-Based Drug Discovery to Versatile Targets · 2020

    No notice
51. 51.
    Iterative computational design and crystallographic screening identifies potent inhibitors targeting the Nsp3 macrodomain of SARS-CoV-2 · 2023

    No notice
52. 52.
    Fragment binding to the Nsp3 macrodomain of SARS-CoV-2 identified through crystallographic screening and computational docking · 2021

    No notice
53. 53.

    Ferla M. P. et al. Fragmenstein: predicting protein-ligand structures of compounds derived from known crystallographic fragment hits using a strict conserved-binding–based methodology. Preprint at ChemRxiv10.26434/chemrxiv-2024-17w01 (2024).

    Not in our library10.1186/s13321-025-00946-0

54. 54.

    Luo, S. et al. Antigen-specific antibody design and optimization with diffusion-based generative models for protein structures. In Advances in Neural Information Processing Systems 9754–9767 (Curran Associates, 2022).

    Not in our library

55. 55.

    Francoeur PG, et al. Three-dimensional convolutional neural networks and a cross-docked data set for structure-based drug design. J. Chem. Inf. Model. 2020;60:4200-4215.

    Not in our library10.1021/acs.jcim.0c00411

56. 56.

    Irwin JJ, et al. ZINC20—a free ultralarge-scale chemical database for ligand discovery. J. Chem. Inf. Model. 2020;60:6065-6073.

    Not in our library10.1021/acs.jcim.0c00675

57. 57.

    Liu Z, et al. Forging the basis for developing protein–ligand interaction scoring functions. Acc. Chem. Res. 2017;50:302-309.

    Not in our library10.1021/acs.accounts.6b00491

58. 58.
    GNINA 1.0: molecular docking with deep learning · 2021

    No notice
59. 59.

    Bickerton GR, Paolini GV, Besnard J, Muresan S, Hopkins AL. Quantifying the chemical beauty of drugs. Nat. Chem. 2012;4:90-98.

    Not in our library10.1038/nchem.1243

60. 60.
    Estimation of synthetic accessibility score of drug-like molecules based on molecular complexity and fragment contributions · 2009

    No notice
61. 61.

    Wildman SA, Crippen GM. Prediction of physicochemical parameters by atomic contributions. J. Chem. Inf. Comput. Sci. 1999;39:868-873.

    Not in our library10.1021/ci990307l

62. 62.

    Lipinski CA, Lombardo F, Dominy BW, Feeney PJ. Experimental and computational approaches to estimate solubility and permeability in drug discovery and development settings. Adv. Drug Deliv. Rev. 2012;64:4-17.

    Not in our library10.1016/j.addr.2012.09.019

63. 63.

    Landrum, G. et al. RDKit: open-source cheminformatics software (RDKit, 2016); https://rdkit.org/

    Not in our library

64. 64.

    Alhossary A, Handoko SD, Mu Y, Kwoh C-K. Fast, accurate, and reliable molecular docking with QuickVina 2. Bioinformatics. 2015;31:2214-2216.

    Not in our library10.1093/bioinformatics/btv082

65. 65.

    Schneuing, A. DiffSBDD datasets. Zenodo10.5281/zenodo.13931612 (2024).

    Not in our library

66. 66.

    Schneuing, A. DiffSBDD molecules. Zenodo10.5281/zenodo.8239058 (2023).

    Not in our library

67. 67.

    Debreczeni, J. E. et al. Crystal structure of the human serine-threonine kinase 16 in complex with staurosporine. Worldwide Protein Data Bank10.2210/pdb2buj/pdb (2005).

    Not in our library

68. 68.

    Sheriff, S. Crystal structure of the mitotic kinesin eg5 in complex with mg-adp and n-(3-aminopropyl)-n-((3-benzyl-5-chloro-4-oxo-3,4-dihydropyrrolo[2,1-f][1,2,4]triazin-2-yl)(cyclopropyl)methyl)-4-methylbenzamide. Worldwide Protein Data Bank10.2210/pdb2GM1/pdb (2006).

    Not in our library

69. 69.

    Chen, H., Zhang, X., Lum, L. & Chen, C. Crystal structure of tankyrase 1 with 355. Worldwide Protein Data Bank10.2210/pdb4TOS/pdb (2015).

    Not in our library

70. 70.

    Sorrell, F. J. et al. Structural Genomics Consortium (SGC) crystal structure of BMP-2-inducible kinase in complex with small molecule AZD-7762. Worldwide Protein Data Bank10.2210/pdb4W9W/pdb (2014).

    Not in our library

71. 71.

    Barone, M., Roske, Y. ENAH EVH1 in complex with ac-[2-cl-F]-[ProM-2]-[ProM-12]-OMe. Worldwide Protein Data Bank10.2210/pdb5NDU/pdb (2018).

    Not in our library

72. 72.

    Correy, G. J., Young, I. D., Thompson, M. C. & Fraser, J. S. PanDDA analysis group deposition—crystal structure of SARS-CoV-2 NSP3 macrodomain in complex with ZINC000000337835. Worldwide Protein Data Bank10.2210/pdb5RSW/pdb (2020).

    Not in our library

73. 73.

    Correy, G. J., Young, I. D., Thompson, M. C. & Fraser, J. S. PanDDA analysis group deposition—crystal structure of SARS-CoV-2 NSP3 macrodomain in complex with ZINC000000000922. Worldwide Protein Data Bank10.2210/pdb5RUE/pdb (2020).

    Not in our library

74. 74.

    Correy, G. J. & Fraser, J.S. PanDDA analysis group deposition—crystal structure of SARS-CoV-2 NSP3 macrodomain in complex with Z4718398539—(R,R) and (S,S) isomers. Worldwide Protein Data Bank10.2210/pdb5SPD/pdb (2022).

    Not in our library

75. 75.

    Chen, P., Lam, K. & Jin, R. Structural basis for recognition of frizzled proteins by Clostridium difficile toxin B. Worldwide Protein Data Bank10.2210/pdb6C0B/pdb (2018).

    Not in our library10.1126/science.aar1999

76. 76.

    Barone, M. & Roske, Y. ENAH EVH1 in complex with ac-[2-cl-F]-[ProM-2]-[ProM-15]-OMe. Worldwide Protein Data Bank10.2210/pdb6RCJ/pdb (2020).

    Not in our library

77. 77.

    Schneuing, A., Harris, C. & Du, Y. DiffSBDD: V0.1. Zenodo10.5281/zenodo.13929691 (2024).

    Not in our library

78. 78.

    Schneuing, A. DiffSBDD models. Zenodo10.5281/zenodo.8183747 (2023).

    Not in our library

Cited by papers we hold 2

• A dual diffusion model enables 3D molecule generation and lead optimization based on target pockets · 2024
  No notice
• Machine Learning-Guided Protein Engineering · 2023
  No notice

Reads

A person counts once a day per paper. We keep no cookie and no address: a reader is a hash that changes every day. Any month under five reads shows no number.