CoChat
Sign Up
PreprintCited by 51Protein Structure and DynamicsChemical Synthesis and AnalysisComputational Drug Discovery Methods

The DOI matches its Crossref record; 1 of the 1 references it cites were found in a registry or an index.

Design of Cyclic Peptides Targeting Protein-Protein Interactions using AlphaFold

Open access
Kosugi, Takatsugu, Ohue, Masahito·2023·International Journal of Molecular Sciences
DOI 10.1101/2023.08.20.554056PMC PMC10487914

Preprint — published as 10.3390/ijms241713257

Open this paper in CoChat — free accountOpen PDF

Abstract

More than 930,000 protein-protein interactions (PPIs) have been identified in recent years, but their physicochemical properties differ from conventional drug targets, complicating the use of conventional small molecules as modalities. Cyclic peptides are a promising modality for targeting protein-protein interactions (PPIs), but it is difficult to predict the structure of a target protein-cyclic peptide complex or to design a cyclic peptide sequence that binds to the target protein using computational methods. Recently, AlphaFold with a cyclic offset has enabled predicting the structure of cyclic peptides, thereby enabling de novo cyclic peptide designs. We developed a cyclic peptide complex offset to enable the structural prediction of target proteins and cyclic peptide complexes and found AlphaFold2 with a cyclic peptide complex offset can predict structures with high accuracy. We also applied the cyclic peptide complex offset to the binder hallucination protocol of AfDesign, a de novo protein design method using AlphaFold, and we could design a high predicted local-distance difference test and lower separated binding energy per unit interface area than the native MDM2/p53 structure. Furthermore, the method was applied to 12 other protein-peptide complexes and one protein-protein complex. Our approach shows that it is possible to design putative cyclic peptide sequences targeting PPI.

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 · 10.1101/2023.08.20.554056

Cites

1 of 1 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1 Schematic diagram of cyclic peptide binder hallucination. (a) Example of relative position encoding of a hypothetical 20-residue protein and a 10-residue peptide. Default complex offset on the left and cyclic peptide complex offset on the right. (b) Schematic showing the optimization of the cyclic peptide binder hallucination. The AF2 output is an all-atom coordinate, two confidence scores (PAE and pLDDT), and a distogram. These are used to define the loss function, which is back-propagated to compute the gradient for the design sequence, then updated and predicted in a loop for optimization. A target protein structure and an initial random sequence are input, and a cyclic peptide is hallucinated to the target protein via the cyclic peptide complex offset. (c) Example sequence logo for a cyclic peptide design targeting MDM2 considering 3D coordinates. The sequence logo represents each column of alignment as a stack of letters, with the height of each letter proportional to the observed frequency of the corresponding amino acid, and the overall height of each stack proportional to the degree of sequence conservation at that location. The width of the letters is proportional to ungaps of each column. Hydrophilic amino acids (R, K, D, E, N, and Q) are shown in blue, neutral amino acids (S, G, H, T, A, and P) are shown in green, and hydrophobic amino acids (Y, V, M, C, L, F, I, and W) are shown in black.
Figure 1 Schematic diagram of cyclic peptide binder hallucination. (a) Example of relative position encoding of a hypothetical 20-residue protein and a 10-residue peptide. Default complex offset on the left and cyclic peptide complex offset on the right. (b) Schematic showing the optimization of the cyclic peptide binder hallucination. The AF2 output is an all-atom coordinate, two confidence scores (PAE and pLDDT), and a distogram. These are used to define the loss function, which is back-propagated to compute the gradient for the design sequence, then updated and predicted in a loop for optimization. A target protein structure and an initial random sequence are input, and a cyclic peptide is hallucinated to the target protein via the cyclic peptide complex offset. (c) Example sequence logo for a cyclic peptide design targeting MDM2 considering 3D coordinates. The sequence logo represents each column of alignment as a stack of letters, with the height of each letter proportional to the observed frequency of the corresponding amino acid, and the overall height of each stack proportional to the degree of sequence conservation at that location. The width of the letters is proportional to ungaps of each column. Hydrophilic amino acids (R, K, D, E, N, and Q) are shown in blue, neutral amino acids (S, G, H, T, A, and P) are shown in green, and hydrophobic amino acids (Y, V, M, C, L, F, I, and W) are shown in black.CC BY · PubMed Central
Figure 2 Comparison of predicted protein–cyclic peptide complex structures of AF2 and ADCP. The left column is the predicted structure of AF2_v2 with the cyclic peptide complex offset. The right column is ADCP local docking structure with the lowest “ref.rmsd” among the top 10 structures. Green shows the predicted structure of the PDB target protein, cyan shows the predicted structure of the AF2 target protein, and magenta shows the predicted structure of the AF2 cyclic peptide. Orange shows the ADCP local docking cyclic peptide structure.
Figure 2 Comparison of predicted protein–cyclic peptide complex structures of AF2 and ADCP. The left column is the predicted structure of AF2_v2 with the cyclic peptide complex offset. The right column is ADCP local docking structure with the lowest “ref.rmsd” among the top 10 structures. Green shows the predicted structure of the PDB target protein, cyan shows the predicted structure of the AF2 target protein, and magenta shows the predicted structure of the AF2 cyclic peptide. Orange shows the ADCP local docking cyclic peptide structure.CC BY · PubMed Central
Figure 3 Comparison of binding energy using the Rosetta interface analyzer score between the native structure and the MDM2-targeted design sequence predicted structure. (a) Scatterplot of binding energy and SASA from the Rosetta interface analyzer. The gray dot presents the native structure value. (b) Scatterplot of binding energy from the Rosetta interface analyzer and pLDDT. The gray line presents the native structure score. (c) Comparison of crystal 1YCR structure and AfDesign sequence structure predicted using ColabFold. The predicted structure of the design sequence had a Rosetta interface analyzer score lower than that of the native structure and had the lowest RMSD_best. Green shows the native target protein structure, cyan shows the predicted target protein structure, and magenta shows the predicted cyclic peptide structure.
Figure 3 Comparison of binding energy using the Rosetta interface analyzer score between the native structure and the MDM2-targeted design sequence predicted structure. (a) Scatterplot of binding energy and SASA from the Rosetta interface analyzer. The gray dot presents the native structure value. (b) Scatterplot of binding energy from the Rosetta interface analyzer and pLDDT. The gray line presents the native structure score. (c) Comparison of crystal 1YCR structure and AfDesign sequence structure predicted using ColabFold. The predicted structure of the design sequence had a Rosetta interface analyzer score lower than that of the native structure and had the lowest RMSD_best. Green shows the native target protein structure, cyan shows the predicted target protein structure, and magenta shows the predicted cyclic peptide structure.CC BY · PubMed Central
Figure 4 Comparison of sequence logos for cyclic and linear peptide designs. The sequence logo represents each column of alignment as a stack of letters, with the height of each letter proportional to the observed frequency of the corresponding amino acid, and the overall height of each stack proportional to the degree of sequence conservation at that location. The width of the letters is proportional to ungaps of each column. (a) Native peptide (p53) sequence logo. (b) The left column shows the sequence logo of cyclic design sequences for 1YCR with each RMSD_best threshold. The right column shows the linear design sequences for 1YCR with each RMSD_best threshold. Hydrophilic amino acids (R, K, D, E, N, and Q) are shown in blue, neutral amino acids (S, G, H, T, A, and P) are shown in green, and hydrophobic amino acids (Y, V, M, C, L, F, I, and W) are shown in black.
Figure 4 Comparison of sequence logos for cyclic and linear peptide designs. The sequence logo represents each column of alignment as a stack of letters, with the height of each letter proportional to the observed frequency of the corresponding amino acid, and the overall height of each stack proportional to the degree of sequence conservation at that location. The width of the letters is proportional to ungaps of each column. (a) Native peptide (p53) sequence logo. (b) The left column shows the sequence logo of cyclic design sequences for 1YCR with each RMSD_best threshold. The right column shows the linear design sequences for 1YCR with each RMSD_best threshold. Hydrophilic amino acids (R, K, D, E, N, and Q) are shown in blue, neutral amino acids (S, G, H, T, A, and P) are shown in green, and hydrophobic amino acids (Y, V, M, C, L, F, I, and W) are shown in black.CC BY · PubMed Central
Figure 5 Comparison of the predicted structures of 12 target proteins and designed cyclic peptides with the native PDB structures. Green shows the predicted structure of the PDB target protein, cyan shows the predicted structure of the AF2 target protein, and magenta shows the predicted structure of the AF2 cyclic peptide.
Figure 5 Comparison of the predicted structures of 12 target proteins and designed cyclic peptides with the native PDB structures. Green shows the predicted structure of the PDB target protein, cyan shows the predicted structure of the AF2 target protein, and magenta shows the predicted structure of the AF2 cyclic peptide.CC BY · PubMed Central
Figure 6 Comparison of binding energy using the Rosetta interface analyzer score between the native structure and PD-L1-targeted design predicted structure. (a) Scatter plot of binding energy and SASA from the Rosetta interface analyzer. The gray dot shows the native structure value. (b) Scatterplot of binding energy from the Rosetta interface analyzer and pLDDT. The gray line shows the native structure score. (c) Comparison of crystal 4ZQK structure and AfDesign sequence structure predicted via ColabFold. The predicted structure with the lowest Rosetta interface analyzer score and pLDDT over 90 is shown. Green shows the native target protein structure, cyan shows the predicted target protein structure, and magenta shows the predicted cyclic peptide structure.
Figure 6 Comparison of binding energy using the Rosetta interface analyzer score between the native structure and PD-L1-targeted design predicted structure. (a) Scatter plot of binding energy and SASA from the Rosetta interface analyzer. The gray dot shows the native structure value. (b) Scatterplot of binding energy from the Rosetta interface analyzer and pLDDT. The gray line shows the native structure score. (c) Comparison of crystal 4ZQK structure and AfDesign sequence structure predicted via ColabFold. The predicted structure with the lowest Rosetta interface analyzer score and pLDDT over 90 is shown. Green shows the native target protein structure, cyan shows the predicted target protein structure, and magenta shows the predicted cyclic peptide structure.CC BY · PubMed Central

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

Research works printed but not linked: 63

  1. 10.1038/468851a · DOI
  2. 10.1002/PRO.3978 · DOI · Paper page
  3. 10.1021/jm010468s · DOI
  4. 10.1038/nrd1343 · DOI · Paper page
  5. 10.1038/nrd1578 · DOI
  6. 10.1146/annurev-pharmtox-011613-140028 · DOI · Paper page
  7. 10.1016/j.tips.2013.04.007 · DOI · Paper page
  8. 2020, Current Challenges and Opportunities in Designing Protein–Protein Interaction Targeted Drugs, Adv. Appl. Bioinform. Chem, 13, 11
  9. 2021, Quantitative Estimate Index for Early-Stage Screening of Compounds Targeting Protein-Protein Interactions, Int. J. Mol. Sci, 22, 10925, 10.3390/ijms222010925 · DOI
  10. 10.1021/jacs.8b13178 · DOI · Paper page
  11. 10.1038/s41573-020-00135-8 · DOI
  12. 2015, Peptide Therapeutics: Targeting the Undruggable Space, Eur. J. Med. Chem, 94, 459, 10.1016/j.ejmech.2015.01.014 · DOI
  13. 2016, Quantifying the Chameleonic Properties of Macrocycles and Other High-Molecular-Weight Drugs, Drug Discov. Today, 21, 712, 10.1016/j.drudis.2016.02.005 · DOI
  14. 2021, Interplay among Conformation, Intramolecular Hydrogen Bonds, and Chameleonicity in the Membrane Permeability and Cyclophilin A Binding of Macrocyclic Peptide Cyclosporin O Derivatives, J. Med. Chem, 64, 8272, 10.1021/acs.jmedchem.1c00211 · DOI
  15. 2021, Large-scale membrane permeability prediction of cyclic peptides crossing a lipid bilayer based on enhanced sampling molecular dynamics simulations, J. Chem. Inf. Model, 61, 3681, 10.1021/acs.jcim.1c00380 · DOI
  16. 2022, Lipid composition is critical for accurate membrane permeability prediction of cyclic peptides by molecular dynamics simulations, J. Chem. Inf. Model, 62, 4549, 10.1021/acs.jcim.2c00931 · DOI
  17. 10.1186/s12929-016-0223-x · DOI
  18. 2021, The RaPID Platform for the Discovery of Pseudo-Natural Macrocyclic Peptides, Acc. Chem. Res, 54, 3604, 10.1021/acs.accounts.1c00391 · DOI
  19. 10.1093/nar/gkp514 · DOI
  20. 2021, Accelerated Antimicrobial Discovery via Deep Generative Models and Molecular Dynamics Simulations, Nat. Biomed. Eng, 5, 613, 10.1038/s41551-021-00689-x · DOI
  21. Cardoso, M.H. ; Orozco, R.Q. ; Rezende, S.B. ; Rodrigues, G. ; Oshiro, K.G.N. ; Cândido, E.S. ; Franco, O.L. Computer-Aided Design of Antimicrobial Peptides: Are We Generating Effective Drug Candidates? Front. Microbiol. 2020, 10, https://doi.org/10.3389/fmicb.2019.03097. · DOI
  22. 10.1021/acs.jcim.9b01014 · DOI
  23. 2020, Recent Advances in the Development of Protein–Protein Interactions Modulators: Mechanisms and Clinical Trials, Signal Transduct. Target Ther, 5, 213, 10.1038/s41392-020-00315-3 · DOI · Paper page
  24. 10.1016/j.drudis.2006.05.012 · DOI
  25. 10.1021/ci200528d · DOI · Paper page
  26. 2022, Virtual Screening Expands the Non-Natural Amino Acid Palette for Peptide Optimization, J. Chem. Inf. Model, 62, 2999, 10.1021/acs.jcim.2c00193 · DOI
  27. 10.1021/jm0306430 · DOI · Paper page
  28. 10.1021/ci400128m · DOI · Paper page
  29. 2017, Docking of Macrocycles: Comparing Rigid and Flexible Docking in Glide, J. Chem. Inf. Model, 57, 190, 10.1021/acs.jcim.6b00443 · DOI
  30. 10.1093/nar/gkv456 · DOI
  31. 2019, CABS-Dock Standalone: A Toolbox for Flexible Protein–Peptide Docking, Bioinformatics, 35, 4170, 10.1093/bioinformatics/btz185 · DOI
  32. 10.1093/bioinformatics/btx216 · DOI
  33. 10.1371/journal.pcbi.1005905 · DOI
  34. 10.1093/nar/gkw329 · DOI
  35. 10.1158/0008-5472.CAN-17-0511 · DOI
  36. 2022, Cyclization and Docking Protocol for Cyclic Peptide–Protein Modeling Using HADDOCK2.4, J. Chem. Theory Comput, 18, 4027, 10.1021/acs.jctc.2c00075 · DOI
  37. 10.1093/bioinformatics/btz459 · DOI
  38. 2022, High Throughput Virtual Screening (HTVS) of Peptide Library: Technological Advancement in Ligand Discovery, Eur. J. Med. Chem, 243, 114766, 10.1016/j.ejmech.2022.114766 · DOI
  39. 10.1126/science.abj8754 · DOI · Paper page
  40. Baek, M. ; Anishchenko, I. ; Humphreys, I.R. ; Cong, Q. ; Baker, D. ; DiMaio, F. Efficient and Accurate Prediction of Protein Structure Using RoseTTAFold2. bioRxiv 2023, https://doi.org/10.1101/2023.05.24.542179. · DOI
  41. Evans, R. ; O’Neill, M. ; Pritzel, A. ; Antropova, N. ; Senior, A. ; Green, T. ; Žídek, A. ; Bates, R. ; Blackwell, S. ; Yim, J. ; et al. Protein Complex Prediction with AlphaFold-Multimer. bioRxiv 2022, https://doi.org/10.1101/2021.10.04.463034. · DOI · Paper page
  42. 2021, Computed Structures of Core Eukaryotic Protein Complexes, Science, 374, eabm4805, 10.1126/science.abm4805 · DOI · Paper page
  43. Gulsevin, A. ; Meiler, J. Benchmarking Peptide Structure Prediction with AlphaFold2. bioRxiv 2022. https://doi.org/10.1101/2022. 02.17.480937. · DOI
  44. 10.1038/s41467-021-27838-9 · DOI · Paper page
  45. 10.1103/PhysRevLett.129.238101 · DOI · Paper page
  46. 2021, De Novo Protein Design by Deep Network Hallucination, Nature, 600, 547, 10.1038/s41586-021-04184-w · DOI · Paper page
  47. 10.1073/pnas.2017228118 · DOI · Paper page
  48. 2023, De Novo Protein Design by Inversion of the AlphaFold Structure Prediction Network, Protein Sci, 32, e4653, 10.1002/pro.4653 · DOI · Paper page
  49. Frank, C. ; Khoshouei, A. ; Stigter, Y. de ; Schiewitz, D. ; Feng, S. ; Ovchinnikov, S. ; Dietz, H. Efficient and Scalable de Novo Protein Design Using a Relaxed Sequence Space. bioRxiv 2023, https://doi.org/10.1101/2023.02.24.529906. · DOI
  50. 2022, Solubility-Aware Protein Binding Peptide Design Using AlphaFold, Biomedicines, 10, 1626, 10.3390/biomedicines10071626 · DOI
  51. Rettie, S.A. ; Campbell, K.V. ; Bera, A.K. ; Kang, A. ; Kozlov, S. ; De La Cruz, J. ; Adebomi, V. ; Zhou, G. ; DiMaio, F. ; Ovchinnikov, S. ; Bhardwaj, G. Cyclic Peptide Structure Prediction and Design Using AlphaFold. bioRxiv 2023, https://doi.org/10.1101/2023.02.25.529 · DOI
  52. 2022, ColabFold - Making Protein Folding Accessible to All, Nat. Methods, 19, 679, 10.1038/s41592-022-01488-1 · DOI · Paper page
  53. 2019, Docking Flexible Cyclic Peptides with AutoDock CrankPep, J. Chem. Theory Comput, 15, 5161, 10.1021/acs.jctc.9b00557 · DOI
  54. 10.1006/jmbi.1998.2401 · DOI
  55. 10.1002/jcc.21256 · DOI
  56. 10.1101/gr.849004 · DOI
  57. 10.1002/pro.2187 · DOI
  58. 10.1093/NAR/GKAB294 · DOI
  59. Banhos Danneskiold-Samøe, N. ; Kavi, D. ; Jude, K.M. ; Nissen, S.B. ; Wat, L.W. ; Coassolo, L. ; Zhao, M. ; Asae Santana-Oikawa, G. ; Broido, B.B. ; Garcia, K.C. ; Svensson, K.J. Rapid and Accurate Deorphanization of Ligand-Receptor Pairs Using AlphaFold. bioRxiv 2023, https://doi.org/10.1101/2023.03.16.531341. · DOI · Paper page
  60. 10.1021/acs.jcim.2c01083 · DOI
  61. Bryant, A. ; Elofsson, A. EvoBind: in Silico Directed Evolution of Peptide Binders with AlphaFold. bioRxiv 2022, https://doi.org/10.1101/2022.07.23.501214. · DOI
  62. Liu, J. ; Chen, Z. ; Li, Y. ; Zhao, W. ; Wu, J. ; Zhang, Z. PD-1/PD-L1 Checkpoint Inhibitors in Tumor Immunotherapy. Front. Pharmacol. 2021, 12, https://doi.org/10.3389/fphar.2021.731798. · DOI
  63. 2021, Rational Design of Potent Peptide Inhibitors of the PD-1:PD-L1 Interaction for Cancer Immunotherapy, J. Am. Chem. Soc, 143, 18536, 10.1021/jacs.1c08132 · DOI · Paper page

Metadata

Authors
Kosugi, T., & Ohue, M.
Affiliations
Tokyo Institute of Technology (JP)
Corresponding author
Ohue, Masahito
ORCID
Venue
International Journal of Molecular Sciences
Publisher
openRxiv
Published
2023-08-21
Volume
24
Issue
17
Pages
13257-13257
ISSN
1422-0067 (ISSN-L)
DOI
10.1101/2023.08.20.554056
PMC
PMC10487914
OpenAlex
W4386222188
Cited by
51
References
1
Citations by year
2026: 14 · 2025: 26 · 2024: 11
FWCI
4.54
Type
journal article
Language
en
Open access
Licence
cc-by
Locations
  • International Journal of Molecular Sciences (publishedVersion, cc-by)
  • bioRxiv (Cold Spring Harbor Laboratory) (acceptedVersion)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
Funders
  • Japan Agency for Medical Research and Development (BINDS JP22ama121026)
  • DeepMind
  • Japan Society for the Promotion of Science (23H04880, JPMJAX20A3, KAKENHI 23H03496, 23K28186, 23H04887)
  • Japan Science and Technology Agency (23H04887, JPMJAX20A3, 23H03496, JP23ama121026, 23H04880, JPMJFR216J)
  • ACT-X
Topics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Protein Structure and Dynamics
  • Physical Sciences › Chemistry › Organic Chemistry › Chemical Synthesis and Analysis
  • Life Sciences › Pharmacology, Toxicology and Pharmaceutics › Drug Discovery › Computational Drug Discovery Methods
Keywords
cyclic peptides, protein-protein interaction, de novo peptide design, AlphaFold2
MeSH terms
Excipients, Orthotic Devices, Peptides, Cyclic, Drug Delivery Systems

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