CoChat
Sign Up
Cited by 14vaccines and immunoinformatics approachesProtein Structure and DynamicsChemical Synthesis and Analysis

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

PANDORA v2.0: Benchmarking peptide-MHC II models and software improvements

Open access
Parizi, Farzaneh M., Marzella, Dario F., Ramakrishnan, Gayatri, ‘t Hoen, Peter A. C., Karimi-Jafari, Mohammad Hossein, Xue, Li C.·2023·Frontiers in Immunology
DOI 10.3389/fimmu.2023.1285899PMC PMC10739464

Published version of a preprint

Open this paper in CoChat — free account

Abstract

T-cell specificity to differentiate between self and non-self relies on T-cell receptor (TCR) recognition of peptides presented by the Major Histocompatibility Complex (MHC). Investigations into the three-dimensional (3D) structures of peptide:MHC (pMHC) complexes have provided valuable insights of MHC functions. Given the limited availability of experimental pMHC structures and considerable diversity of peptides and MHC alleles, it calls for the development of efficient and reliable computational approaches for modeling pMHC structures. Here we present an update of PANDORA and the systematic evaluation of its performance in modelling 3D structures of pMHC class II complexes (pMHC-II), which play a key role in the cancer immune response. PANDORA is a modelling software that can build low-energy models in a few minutes by restraining peptide residues inside the MHC-II binding groove. We benchmarked PANDORA on 136 experimentally determined pMHC-II structures covering 44 unique αβ chain pairs. Our pipeline achieves a median backbone Ligand-Root Mean Squared Deviation (L-RMSD) of 0.42 Å on the binding core and 0.88 Å on the whole peptide for the benchmark dataset. We incorporated software improvements to make PANDORA a pan-allele framework and improved the user interface and software quality. Its computational efficiency allows enriching the wealth of pMHC binding affinity and mass spectrometry data with 3D models. These models can be used as a starting point for molecular dynamics simulations or structure-boosted deep learning algorithms to identify MHC-binding peptides. PANDORA is available as a Python package through Conda or as a source installation at https://github.com/X-lab-3D/PANDORA .

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 · 10.3389/fimmu.2023.1285899

Cites

3 of 3 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1 Overview of the pMHC-II complex (A) Representation of an MHC-II molecule by its accessible surface area, visualized with Protein Imager (19). MHC-II consists of an α-chain (light blue) and a β-chain (dark blue). Shown are the four characteristic pockets in the binding groove (P1, P4, P6, and P9), occupied by the corresponding peptide (orange) anchor residues. As modeling restraints (yellow), PANDORA uses the atomic contacts between the peptide anchor residues and the MHC-II pockets. (B) Cartoon representation of pMHC-II. The peptide binding groove consists of two α-helices on a floor of β-sheets, in which the peptide resides (PDB ID: 1DLH).
Figure 1 Overview of the pMHC-II complex (A) Representation of an MHC-II molecule by its accessible surface area, visualized with Protein Imager (19). MHC-II consists of an α-chain (light blue) and a β-chain (dark blue). Shown are the four characteristic pockets in the binding groove (P1, P4, P6, and P9), occupied by the corresponding peptide (orange) anchor residues. As modeling restraints (yellow), PANDORA uses the atomic contacts between the peptide anchor residues and the MHC-II pockets. (B) Cartoon representation of pMHC-II. The peptide binding groove consists of two α-helices on a floor of β-sheets, in which the peptide resides (PDB ID: 1DLH).CC BY · PubMed Central
Figure 2 Overview of the PANDORA pMHC modelling framework. PANDORA as an integrative modelling protocol, leverages two domain knowledge aspects: the highly conserved nature of MHC structures and the binding of peptides to MHC pockets with anchor residues. PANDORA takes the sequence information of a target peptide and MHC as input and selects a template pMHC structure from a template set based on sequence similarity. The target peptide core is superposed onto the template peptide core. In flexible mode, it applies distance restraints for anchor residues. The framework performs loop modeling of flexible regions and energy minimization of pMHC-II conformations. Conformations are ranked to select those resembling the near-native conformation.
Figure 2 Overview of the PANDORA pMHC modelling framework. PANDORA as an integrative modelling protocol, leverages two domain knowledge aspects: the highly conserved nature of MHC structures and the binding of peptides to MHC pockets with anchor residues. PANDORA takes the sequence information of a target peptide and MHC as input and selects a template pMHC structure from a template set based on sequence similarity. The target peptide core is superposed onto the template peptide core. In flexible mode, it applies distance restraints for anchor residues. The framework performs loop modeling of flexible regions and energy minimization of pMHC-II conformations. Conformations are ranked to select those resembling the near-native conformation.CC BY · PubMed Central
Figure 3 Benchmark results on reproducing 136 pMHC-II complexes with X-ray structures. (A) Sampling performance of the PANDORA benchmark experiment. The conformation with the lowest RMSD was chosen as the best RMSD model. A circular bar plot grouped based on peptide length (represented by the numbers in the inner circle) reports the lowest backbone L-RMSD (Y-axis) for the whole peptide (navy) and binding core (yellow). (B) An example of an average-quality 3D model generated by PANDORA. The target peptide (PDB ID: 4I5B) is marked in green; the template structure (PDB ID: 2OJE) is marked in magenta; and the PANDORA model structure (best conformation among the top 5 ranked) is marked in darkblue. (C, D) Histogram of the lowest backbone L-RMSD models in the peptide binding core vs. the whole peptide. (E, F) Complete performance of PANDORA (modeling + scoring). Histogram for the top-ranked models by PANDORA in terms of backbone L-RMSD on the peptide binding core vs. the whole peptide.
Figure 3 Benchmark results on reproducing 136 pMHC-II complexes with X-ray structures. (A) Sampling performance of the PANDORA benchmark experiment. The conformation with the lowest RMSD was chosen as the best RMSD model. A circular bar plot grouped based on peptide length (represented by the numbers in the inner circle) reports the lowest backbone L-RMSD (Y-axis) for the whole peptide (navy) and binding core (yellow). (B) An example of an average-quality 3D model generated by PANDORA. The target peptide (PDB ID: 4I5B) is marked in green; the template structure (PDB ID: 2OJE) is marked in magenta; and the PANDORA model structure (best conformation among the top 5 ranked) is marked in darkblue. (C, D) Histogram of the lowest backbone L-RMSD models in the peptide binding core vs. the whole peptide. (E, F) Complete performance of PANDORA (modeling + scoring). Histogram for the top-ranked models by PANDORA in terms of backbone L-RMSD on the peptide binding core vs. the whole peptide.CC BY · PubMed Central
Figure 4 Comparison of PANDORA and AlphaFold in reproducing pMHC-II complexes. The peptide conformations are colored as follows: reference PDB in green, AlphaFold model in magenta, and PANDORA model in blue. The AlphaFold models were generated by Colab-fold using template-based modelling (default 5 top models generated). The presented model for AlphaFold and PANDORA is the best model (lowest L-RMSD model) among the top-5 ranked models. Two overviews of the peptide backbone conformation from two different angles are shown for each case (anchor residues are shown with side-chains). Overall, AlphaFold models have quite good backbone predictions (probably due to the usage of templates), but in 2 out of 4 cases, the peptide core conformations are shifted. (A) 24-mer peptide binding to H2-AB1*01 (PDB ID: 3C5Z); however, the predicted core conformation generated by AlphaFold is shifted by 3 residues; (B) 13-mer peptide binding to HLA-DRA*0101, HLA-DRB1*0401 (PDB ID: 1J8H) the binding core is accurately identified by AlphaFold (C) 12-mer peptide binding to HLA-DQA1*0201, HLA-DQB1*0201 (PDB ID: 6PX6); however, the predicted AlphaFold binding core conformation is shifted by 4 residues; (D) 13-mer peptide binding to H2-AB1*01 (PDB ID: 4P23) and the binding core is accurately identified by AlphaFold. Considering that (A, B, D) were already in AlphaFold training, it is noteworthy that A is predicted with an incorrect binding core.
Figure 4 Comparison of PANDORA and AlphaFold in reproducing pMHC-II complexes. The peptide conformations are colored as follows: reference PDB in green, AlphaFold model in magenta, and PANDORA model in blue. The AlphaFold models were generated by Colab-fold using template-based modelling (default 5 top models generated). The presented model for AlphaFold and PANDORA is the best model (lowest L-RMSD model) among the top-5 ranked models. Two overviews of the peptide backbone conformation from two different angles are shown for each case (anchor residues are shown with side-chains). Overall, AlphaFold models have quite good backbone predictions (probably due to the usage of templates), but in 2 out of 4 cases, the peptide core conformations are shifted. (A) 24-mer peptide binding to H2-AB1*01 (PDB ID: 3C5Z); however, the predicted core conformation generated by AlphaFold is shifted by 3 residues; (B) 13-mer peptide binding to HLA-DRA*0101, HLA-DRB1*0401 (PDB ID: 1J8H) the binding core is accurately identified by AlphaFold (C) 12-mer peptide binding to HLA-DQA1*0201, HLA-DQB1*0201 (PDB ID: 6PX6); however, the predicted AlphaFold binding core conformation is shifted by 4 residues; (D) 13-mer peptide binding to H2-AB1*01 (PDB ID: 4P23) and the binding core is accurately identified by AlphaFold. Considering that (A, B, D) were already in AlphaFold training, it is noteworthy that A is predicted with an incorrect binding core.CC BY · PubMed Central
Figure 5 The effect of modelling with a different template allele-type and its effect on performance. Given two allele-type for each α- and β-chains of the template and target pMHC-II, 4 different scenarios are compared in each box-plot column; 1) Both allele-type is the same for template and target (blue); 2) Only the Alpha allele-type the same (green); 3) Only Beta allele-type the same (orange); and 4) both chain allele-types are different (pink). (A) Best L-RMSD models and (B) Top-ranked models using scoring.
Figure 5 The effect of modelling with a different template allele-type and its effect on performance. Given two allele-type for each α- and β-chains of the template and target pMHC-II, 4 different scenarios are compared in each box-plot column; 1) Both allele-type is the same for template and target (blue); 2) Only the Alpha allele-type the same (green); 3) Only Beta allele-type the same (orange); and 4) both chain allele-types are different (pink). (A) Best L-RMSD models and (B) Top-ranked models using scoring.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: 49

  1. Janeway, 2001, Immunobiology: the immune system in health and disease, 732
  2. Blum, 2013, Pathways of antigen processing, Annu Rev Immunol, 31, 10.1146/annurev-immunol-032712-095910 · DOI
  3. Madden, 1995, The three-dimensional structure of peptide-MHC complexes, Annu Rev Immunol, 13, 587, 10.1146/annurev.iy.13.040195.003103 · DOI
  4. Abualrous, 2023, MHC-II dynamics are maintained in HLA-DR allotypes to ensure catalyzed peptide exchange, Nat Chem Biol, 19, 1, 10.1038/s41589-023-01316-3 · DOI
  5. Thomas, 2017, Structure of the TAPBPR–MHC I complex defines the mechanism of peptide loading and editing, Science, 358, 10.1126/science.aao6001 · DOI
  6. Zhang, 1998, Structural principles that govern the peptide-binding motifs of class I MHC molecules, JMB, 281, 10.1006/jmbi.1998.1982 · DOI
  7. Sinigaglia, 1994, Defining rules for the peptide-MHC class II interaction, Curr Opin Immunol, 6, 10.1016/0952-7915(94)90033-7 · DOI
  8. Ferrante, 2007, Cooperativity of hydrophobic anchor interactions: evidence for epitope selection by MHC class II as a folding process, J Immunol, 178, 10.4049/jimmunol.178.11.7181 · DOI
  9. Murthy, 1997, The class II MHC protein HLA-DR1 in complex with an endogenous peptide: implications for the structural basis of the specificity of peptide binding, Structure, 5, 10.1016/S0969-2126(97)00288-8 · DOI
  10. Wucherpfennig, 2001, Insights into autoimmunity gained from structural analysis of MHC–peptide complexes, Curr Opin Immunol, 13, 10.1016/S0952-7915(01)00274-6 · DOI
  11. Madura, 2015, Structural basis for ineffective T-cell responses to MHC anchor residue-improved “heteroclitic” peptides: Molecular immunology, Eur J Immunol, 45, 10.1002/eji.201445114 · DOI
  12. Smith, 2021, Structurally silent peptide anchor modifications allosterically modulate T cell recognition in a receptor-dependent manner, Proc Natl Acad Sci USA, 118, e2018125118, 10.1073/pnas.2018125118 · DOI
  13. Saotome, 2023, Structural analysis of cancer-relevant TCR-CD3 and peptide-MHC complexes by cryoEM, Nat Commun, 14, 2401, 10.1038/s41467-023-37532-7 · DOI
  14. Sušac, 2022, Structure of a fully assembled tumor-specific T cell receptor ligated by pMHC, Cell, 185, 3201, 10.1016/j.cell.2022.07.010 · DOI
  15. Sahin, 2017, Personalized RNA mutanome vaccines mobilize poly-specific therapeutic immunity against cancer, Nature, 547, 10.1038/nature23003 · DOI · Paper page
  16. Ott, 2017, An immunogenic personal neoantigen vaccine for patients with melanoma, Nature, 547, 10.1038/nature22991 · DOI · Paper page
  17. Kreiter, 2015, Mutant MHC class II epitopes drive therapeutic immune responses to cancer, Nature, 520, 10.1038/nature14426 · DOI · Paper page
  18. Tomasello, 2020, The Protein Imager: a full-featured online molecular viewer interface with server-side HQ-rendering capabilities, Bioinformatics, 36, 10.1093/bioinformatics/btaa009 · DOI
  19. Jardetzky, 1996, Crystallographic analysis of endogenous peptides associated with HLA-DR1 suggests a common, polyproline II-like conformation for bound peptides, Proc Natl Acad Sci USA, 93, 10.1073/pnas.93.2.734 · DOI
  20. Wieczorek, 2017, Major histocompatibility complex (MHC) class I and MHC class II proteins: conformational plasticity in antigen presentation, Front Immunol, 8, 10.3389/fimmu.2017.00292 · DOI
  21. Hammer, 1994, Precise prediction of major histocompatibility complex class II-peptide interaction based on peptide side chain scanning, J Exp Med, 180, 10.1084/jem.180.6.2353 · DOI
  22. Stern, 1994, Crystal structure of the human class II MHC protein HLA-DR1 complexed with an influenza virus peptide, Nature, 368, 10.1038/368215a0 · DOI
  23. Brown, 1993, Three-dimensional structure of the human class II histocompatibility antigen HLA-DR1, Nature, 364, 10.1038/364033a0 · DOI
  24. Kulski, 2022, Human leukocyte antigen super-locus: nexus of genomic supergenes, SNPs, indels, transcripts, and haplotypes, Hum Genome Var, 9, 49, 10.1038/s41439-022-00226-5 · DOI
  25. Barker, 2023, The IPD-IMGT/HLA database, Nucleic Acids Res, 51, 10.1093/nar/gkac1011 · DOI
  26. Berman, 2000, The protein data bank, Nucleic Acids Res, 28, 10.1093/nar/28.1.235 · DOI · Paper page
  27. Patronov, 2011, Peptide binding prediction for the human class II MHC allele HLA-DP2: a molecular docking approach, BMC Struct Biol, 11, 32, 10.1186/1472-6807-11-32 · DOI
  28. Patronov, 2014, Histidine Hydrogen Bonding in MHC at pH 5 and pH 7 Modeled by Molecular Docking and Molecular Dynamics Simulations, CAD, 10, 10.2174/15734099113096660050 · DOI
  29. Bordner, 2010, Towards universal structure-based prediction of class II MHC epitopes for diverse allotypes, PloS One, 5, 10.1371/journal.pone.0014383 · DOI
  30. Tong, 2004, Modeling the structure of bound peptide ligands to major histocompatibility complex, Protein Sci, 13, 10.1110/ps.04631204 · DOI
  31. Khan, 2010, pDOCK: a new technique for rapid and accurate docking of peptide ligands to Major Histocompatibility Complexes, Immunome Res, 6, 10.1186/1745-7580-6-S1-S2 · DOI
  32. Atanasova, 2013, EpiDOCK: a molecular docking-based tool for MHC class II binding prediction, PEDS, 26, 10.1093/protein/gzt018 · DOI
  33. Racle, 2019, Robust prediction of HLA class II epitopes by deep motif deconvolution of immunopeptidomes, Nat Biotechnol, 37, 10.1038/s41587-019-0289-6 · DOI
  34. Liu, 2022, DeepSeqPanII: an interpretable recurrent neural network model with attention mechanism for peptide-HLA class II binding prediction, IEEE/ACM Trans Comput Biol Bioinf., 19, 10.1109/TCBB.2021.3074927 · DOI
  35. Reynisson, 2020, NetMHCpan-4.1 and NetMHCIIpan-4.0: improved predictions of MHC antigen presentation by concurrent motif deconvolution and integration of MS MHC eluted ligand data, NAR, 48, 10.1093/nar/gkaa379 · DOI · Paper page
  36. Marzella, 2022, PANDORA: A fast, anchor-restrained modelling protocol for peptide: MHC complexes, Front Immunol, 13, 10.3389/fimmu.2022.878762 · DOI
  37. Marzella, 2023, The PANDORA software for anchor-restrained peptide:MHC modeling, Methods Mol Biol, 2673, 10.1007/978-1-0716-3239-0_18 · DOI
  38. Ehrenmann, 2010, IMGT/3Dstructure-DB and IMGT/DomainGapAlign: a database and a tool for immunoglobulins or antibodies, T cell receptors, MHC, IgSF and MhcSF, Nucleic Acids Res, 38, 10.1093/nar/gkp946 · DOI
  39. Webb, 2016, Comparative protein structure modeling using MODELLER, Curr Protoc Bioinf, 54, 1, 10.1002/prot.21804 · DOI
  40. Lensink, 2007, Docking and scoring protein complexes: CAPRI 3rd Edition, Proteins, 69, 10.1002/prot.21804 · DOI
  41. Mirdita, 2022, ColabFold: making protein folding accessible to all, Nat Methods, 19, 10.1038/s41592-022-01488-1 · DOI · Paper page
  42. Evans, 2021, Protein complex prediction with AlphaFold-Multimer, Bioinformatics;, 4, 10.1101/2021.10.04.463034 · DOI · Paper page
  43. Carson, 1997, T cell receptor recognition of MHC class II–bound peptide flanking residues enhances immunogenicity and results in altered TCR V region usage, Immunity, 7, 10.1016/S1074-7613(00)80360-X · DOI
  44. Vignali, 1994, Amino acid residues that flank core peptide epitopes and the extracellular domains of CD4 modulate differential signaling through the T cell receptor, JEM, 179, 10.1084/jem.179.6.1945 · DOI
  45. Muller, 1996, Activation of T cells by the ragged tail of MHC class II-presented peptides of the measles virus fusion protein, Int Immunol, 8, 10.1093/intimm/8.4.445 · DOI
  46. Zavala-Ruiz, 2004, A hairpin turn in a class II MHC-bound peptide orients residues outside the binding groove for T cell recognition, Proc Natl Acad Sci USA, 101, 10.1073/pnas.0403371101 · DOI
  47. Levy, 2023, Large-scale immunopeptidome analysis reveals recurrent posttranslational splicing of cancer- and immune-associated genes, MCP, 22, 100519, 10.1016/j.mcpro.2023.100519 · DOI
  48. Bloodworth, 2022, Rosetta FlexPepDock to predict peptide-MHC binding: An approach for non-canonical amino acids, PloS One, 17, 10.1371/journal.pone.0275759 · DOI
  49. Sandalova, 2022, Structural aspects of chemical modifications in the MHC-restricted immunopeptidome; Implications for immune recognition, Front Chem, 10, 10.3389/fchem.2022.861609 · DOI

Metadata

Authors
Parizi, F. M., Marzella, D. F., Ramakrishnan, G., ‘t Hoen, P. A. C., Karimi-Jafari, M. H., & Xue, L. C.
Affiliations
  • University of Tehran (IR)
  • Radboud University Medical Center (NL)
Corresponding author
Xue, Li C.
ORCID
Venue
Frontiers in Immunology
Publisher
Frontiers Media SA
Published
2023-12-08
Volume
14
Pages
1285899-1285899
ISSN
1664-3224 (ISSN-L)
DOI
10.3389/fimmu.2023.1285899
PMC
PMC10739464
OpenAlex
W4389479390
Cited by
14
References
3
Citations by year
2026: 3 · 2025: 4 · 2024: 6 · 2023: 1
FWCI
1.24
Type
journal article
Language
en
Open access
Licence
Locations
  • Frontiers in Immunology (publishedVersion, cc-by)
  • bioRxiv (Cold Spring Harbor Laboratory) (acceptedVersion)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • Radboud Repository (Radboud University) (submittedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
  • Radboud Universitair Medisch Centrum
  • European Regional Development Fund
Topics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › vaccines and immunoinformatics approaches
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Protein Structure and Dynamics
  • Physical Sciences › Chemistry › Organic Chemistry › Chemical Synthesis and Analysis
Keywords
Pandora, peptide-MHC class II complexes, T cell receptor recognition, benchmarking, cancer immune response
MeSH terms
Histocompatibility Antigens, Major Histocompatibility Complex, Peptides, Software, Benchmarking

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