CoChat
Sign Up
PreprintCited by 22Computational Drug Discovery MethodsProtein Structure and DynamicsProtein Degradation and Inhibitors

The DOI matches its Crossref record; 3 of the 3 references it cites were found in a registry or an index.

Atom-level generative foundation model for molecular interaction with pockets

Open access
Peng, Xingang, Guo, Fenglin, Guo, Ruihan, Sun, Jiayu, Guan, Jiaqi, Jia, Yinjun, Xu, Yan, Huang, Yanwen, Zhang, Muhan, Peng, Jian, Wang, Xinquan, Han, Chuanhui, Wang, Zihua, Ma, Jianzhu·2024·Cell
DOI 10.1101/2024.10.17.618827

Preprint — published as 10.1016/j.cell.2026.01.003 · Paper page

Open this paper in CoChat — free account

Abstract

Understanding molecular interactions is essential to structural biology and drug discovery. Despite the progress of AI models in revealing and exploiting the interaction mechanisms for various applications, they are predominantly tailored to specific tasks without fully exploiting the underlying transferability across molecular data and tasks. Here, we present PocketXMol, an atom-level generative foundation model to decipher fundamental atomic interactions for general protein-pocket-interacting molecular tasks. It adopts a novel unified generative framework with an innovative task prompt mechanism and an exclusive atom-level representation, making it applicable to diverse tasks covering structure prediction and design of small molecules and peptides, without requiring fine-tuning. PocketXMol was compared to 55 baseline models across 13 typical tasks, achieving state-of-the-art performance in 11 tasks and remaining competitive in the others. We successfully utilized PocketXMol to design novel small molecules that inhibit caspase-9 with efficacy comparable to that of commercial pan-caspase inhibitors. Furthermore, we employed PocketXMol to design PD-L1-binding peptides, demonstrating a success rate substantially higher than random library screening. Three representative peptides underwent further experiments, which validated the cellular specificity and confirmed their potential for molecular probing and therapeutics. PocketXMol presents a powerful and versatile tool with promising prospects for future applications and will have a profound impact on AI-aided drug discovery.

Overview

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1101/2024.10.17.618827

Cites

3 of 3 references verified0 citation errors · 0 not checkable

Statistics

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Flagged first, then by citations

Research works printed but not linked: 106

  1. 2000, Structural symmetry and protein function, Annu. review biophysics biomolecular structure, 29, 105, 10.1146/annurev.biophys.29.1.105 · DOI · Paper page
  2. Bergethon, P. R. The physical basis of biochemistry: the foundations of molecular biophysics (Springer Science & Business Media, 2013).
  3. 10.1101/2024.07.01.600583 · DOI
  4. 2022, Language models of protein sequences at the scale of evolution enable accurate structure prediction, BioRxiv, 2022, 500902
  5. 10.1038/s41586-024-07487-w · DOI · Paper page
  6. 10.1126/science.ade2574 · DOI · Paper page
  7. 10.1038/s41580-019-0163-x · DOI · Paper page
  8. Wu, R. et al. High-resolution de novo structure prediction from primary sequence. BioRxiv 2022–07 (2022). · DOI · Paper page
  9. 10.1126/science.adl2528 · DOI · Paper page
  10. 10.1126/science.add2187 · DOI · Paper page
  11. 10.1038/s41586-023-06415-8 · DOI · Paper page
  12. 10.1038/s41586-023-06728-8 · DOI · Paper page
  13. 10.1038/s41392-022-00904-4 · DOI · Paper page
  14. 10.48550/ARXIV.2306.11768 · DOI
  15. 1984, The characterization of atomic interactions. The J, chemical physics, 80, 1943
  16. 10.48550/ARXIV.2203.02155 · DOI · Paper page
  17. 10.48550/ARXIV.2204.06125 · DOI · Paper page
  18. 10.48550/ARXIV.2112.10752 · DOI · Paper page
  19. 10.1038/s41586-023-06139-9 · DOI · Paper page
  20. 10.1101/2023.05.29.542705 · DOI · Paper page
  21. 10.1038/s41592-024-02201-0 · DOI · Paper page
  22. 10.1038/s41586-024-07618-3 · DOI · Paper page
  23. 10.1038/s41586-024-07441-w · DOI · Paper page
  24. Bommasani, R. , et al. On the opportunities and risks of foundation models. ArXiv (2021).
  25. 10.1038/s41467-022-32007-7 · DOI · Paper page
  26. Liu, S. , et al. Chatgpt-powered conversational drug editing using retrieval and domain feedback. arXiv preprint arXiv:2305.18090 (2023). · arXiv
  27. Ho, J. , Jain, A. & Abbeel, P . Denoising diffusion probabilistic models. In Proceedings of the 34th International Conference on Neural Information Processing Systems, NIPS’20 (Curran Associates Inc., Red Hook, NY, USA, 2020).
  28. Song, Y. , et al. Score-based generative modeling through stochastic differential equations. In International Conference on Learning Representations (2021).
  29. Corso, G. , Stärk, H. , Jing, B. , Barzilay, R. & Jaakkola, T. S . Diffdock: Diffusion steps, twists, and turns for molecular docking. In The Eleventh International Conference on Learning Representations (2023).
  30. Guan, J. , et al. 3d equivariant diffusion for target-aware molecule generation and affinity prediction. In The Eleventh International Conference on Learning Representations (2023).
  31. 10.1073/pnas.2311500121 · DOI
  32. 10.1093/bioinformatics/btu626 · DOI
  33. 10.1093/nar/gkm911 · DOI
  34. 10.1021/acs.jcim.0c00411 · DOI
  35. 10.1093/bioinformatics/bty579 · DOI
  36. 10.1093/nar/gkr777 · DOI · Paper page
  37. 10.1021/acs.jcim.0c00675 · DOI
  38. 10.1038/s41597-022-01288-4 · DOI
  39. Grambow, C. A. , Weir, H. , Cunningham, C. N. , Biancalani, T. & Chuang, K. V. Cremp: Conformer-rotamer ensembles of macrocyclic peptides for machine learning. arXiv preprint arXiv:2305.08057 (2023). · DOI
  40. Zhou, G. , et al. Uni-mol: A universal 3d molecular representation learning framework. In The Eleventh International Conference on Learning Representations (2023). · DOI
  41. Ye, G. , et al. Drugassist: A large language model for molecule optimization. arXiv preprint arXiv:2401.10334 (2023). · DOI
  42. 10.1186/s13321-024-00829-w · DOI
  43. Peng, X. , et al. Pocket2mol: Efficient molecular sampling based on 3d protein pockets. In International Conference on Machine Learning (2022).
  44. Masuda, T. , Ragoza, M. & Koes, D. R . Generating 3d molecular structures conditional on a receptor binding site with deep generative models. arXiv preprint arXiv:2010.14442 (2020). · arXiv
  45. 2021, Advances in Neural Information Processing Systems, 34, 6229
  46. Guan, J. et al. DecompDiff: Diffusion models with decomposed priors for structure-based drug design. In Krause, A. et al. (eds.) Proceedings of the 40th International Conference on Machine Learning, vol. 202 of Proceedings of Machine Learning Research, 11827–11846 (PMLR, 2023).
  47. Huang, Z. , et al. Protein-ligand interaction prior for binding-aware 3d molecule diffusion models. In The Twelfth International Conference on Learning Representations (2024).
  48. 2024, Aligning target-aware molecule diffusion models with exact energy optimization, Adv. Neural Inf. Process. Syst, 37, 44040
  49. 10.1021/acs.jcim.1c00203 · DOI
  50. 10.1039/d3sc04185a · DOI
  51. 10.48550/ARXIV.2308.07413 · DOI
  52. 10.1101/2021.10.04.463034 · DOI · Paper page
  53. 10.1021/acs.jctc.7b00125 · DOI · Paper page
  54. 10.48550/ARXIV.2203.17003 · DOI
  55. Peng, X. , Guan, J. , Liu, Q. & Ma, J. MolDiff: Addressing the atom-bond inconsistency problem in 3D molecule diffusion generation. In Krause, A. et al. (eds.) Proceedings of the 40th International Conference on Machine Learning, vol. 202 of Proceedings of Machine Learning Research, 27611–27629 (PMLR, 2023).
  56. 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).
  57. 10.1109/tnnls.2024.3416328 · DOI
  58. 10.1038/s42004-024-01233-z · DOI
  59. 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. ChemRxiv (2024). · DOI
  60. 10.1021/acs.jmedchem.8b01827 · DOI
  61. 10.1021/acs.jcim.9b01120 · DOI
  62. 10.1038/s41573-021-00371-6 · DOI · Paper page
  63. 10.48550/ARXIV.2205.07309 · DOI
  64. Igashov, I. , et al. Equivariant 3d-conditional diffusion models for molecular linker design. arXiv preprint arXiv:2210.05274 (2022). · arXiv
  65. Guan, J. , et al. Linkernet: Fragment poses and linker co-design with 3d equivariant diffusion. In Thirty-seventh Conference on Neural Information Processing Systems (2023).
  66. 10.1093/nar/gkac946 · DOI
  67. 10.1002/cmdc.200800178 · DOI
  68. 2000, Acd/log p method description, Perspectives drug discovery design, 19, 99, 10.1023/A:1008719622770 · DOI
  69. 10.1038/s41590-020-0641-5 · DOI
  70. 10.1126/sciimmunol.abc6998 · DOI
  71. 10.1016/j.ccell.2019.04.003 · DOI
  72. 10.1093/nar/gky1033 · DOI · Paper page
  73. 10.1371/journal.pbio.0030183 · DOI
  74. 10.1002/pro.3330 · DOI · Paper page
  75. 2023, Design of cyclic peptides targeting protein–protein interactions using alphafold, Int. J. Mol. Sci, 24, 13257, 10.3390/ijms241713257 · DOI
  76. 2024, Aaindexnc: Estimating the physicochemical properties of non-canonical amino acids, including those derived from the pdb and pdbechem databank, Int. J. Mol. Sci, 25, 12555, 10.3390/ijms252312555 · DOI
  77. 10.1093/nar/gki387 · DOI · Paper page
  78. 10.1038/s42256-021-00409-9 · DOI
  79. 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).
  80. 10.1101/2022.06.06.495043 · DOI
  81. Alcaide, E. , et al. Uni-mol docking v2: Towards realistic and accurate binding pose prediction. arXiv preprint arXiv:2405.11769 (2024). · arXiv
  82. 10.1002/prot.10465 · DOI
  83. 10.1101/2023.12.13.571408 · DOI · Paper page
  84. 2021, Advances in Neural Information Processing Systems, 34, 19784
  85. Shi, C. , Luo, S. , Xu, M. & Tang, J . Learning gradient fields for molecular conformation generation. In International Conference on Machine Learning (2021).
  86. Qiao, Z. , Nie, W. , Vahdat, A. , Miller III, T. F. & Anandkumar, A . State-specific protein– ligand complex structure prediction with a multiscale deep generative model. Nat . Mach. Intell. 1–14 (2024). · DOI
  87. 10.3389/fbinf.2022.959160 · DOI · Paper page
  88. Tsaban, T. et al. Harnessing protein folding neural networks for peptide–protein docking. Nat . communications 13, 176 (2022). · DOI · Paper page
  89. 10.1038/s41467-021-25772-4 · DOI
  90. 10.1371/journal.pone.0161879 · DOI · Paper page
  91. 10.1093/nar/gkad630 · DOI
  92. 10.1093/gpbjnl/qzae001 · DOI
  93. 10.1016/j.str.2015.03.010 · DOI
  94. 10.1016/j.jmb.2010.07.021 · DOI
  95. 10.1371/journal.pcbi.1002288 · DOI
  96. 10.1038/s42003-023-04927-7 · DOI
  97. 10.1093/bib/bbae215 · DOI
  98. 10.1101/2023.02.25.529956 · DOI
  99. 10.1002/pro.4653 · DOI · Paper page
  100. 2021, Diffusion models beat gans on image synthesis, Adv. Neural Inf. Process. Syst, 34, 8780
  101. 10.48550/arXiv.2208.09392 · DOI
  102. 10.48550/arXiv.2209.05442 · DOI
  103. Leach, A. , Schmon, S. M. , Degiacomi, M. T. & Willcocks, C. G . Denoising diffusion probabilistic models on SO(3) for rotational alignment. In ICLR 2022 Workshop on Geometrical and Topological Representation Learning (2022).
  104. Loshchilov, I. & Hutter, F . Decoupled weight decay regularization. In International Conference on Learning Representations (2019).
  105. 2015, Microarray based screening of peptide nano probes for her2 positive tumor. Anal, chemistry, 87, 8367
  106. 10.1093/bioinformatics/bty1057 · DOI · Paper page

Metadata

Authors
Peng, X., Guo, F., Guo, R., Sun, J., Guan, J., Jia, Y., Xu, Y., Huang, Y., Zhang, M., Peng, J., Wang, X., Han, C., Wang, Z., & Ma, J.
Affiliations
  • Peking University (CN)
  • Tsinghua University (CN)
  • University of Illinois Urbana-Champaign (US)
  • Center for Life Sciences (CN)
  • Fujian Medical University (CN)
Corresponding author
  • Wang, Xinquan
  • Wang, Zihua
  • Ma, Jianzhu
ORCID
Venue
Cell
Publisher
openRxiv
Published
2024-10-21
Volume
189
Issue
7
Pages
1904-1922.e28
ISSN
0092-8674 (ISSN-L)
DOI
10.1101/2024.10.17.618827
OpenAlex
W7130380853
Cited by
22
References
3
Citations by year
2026: 21 · 2025: 1
FWCI
76.24
Type
journal article
Language
en
Open access
Licence
Locations
  • Cell (publishedVersion, cc-by)
  • bioRxiv (Cold Spring Harbor Laboratory) (acceptedVersion, cc-by)
  • PubMed (publishedVersion)
Funders
National Natural Science Foundation of China
Topics
  • Life Sciences › Pharmacology, Toxicology and Pharmaceutics › Drug Discovery › Computational Drug Discovery Methods
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Protein Structure and Dynamics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Protein Degradation and Inhibitors
Keywords
3D molecular generation, peptide design, structure prediction
MeSH terms
Humans, Models, Molecular, Peptides, Protein Binding, Drug Design, Caspase 9, Small Molecule Libraries, Drug Discovery, Caspase Inhibitors, Molecular Docking Simulation

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review