CoChat
Sign Up
PreprintCited by 21Protein Structure and DynamicsMachine Learning in BioinformaticsEnzyme Structure and Function

The DOI matches its Crossref record; 2 of the 2 references it cites were found in a registry or an index.

Exploring the alternative conformation of a known protein structure based on contact map prediction

Open access
Li, Jiaxuan, Wang, Lei, Zhu, Zefeng, Song, Chen·2022·Journal of Chemical Information and Modeling
DOI 10.1101/2022.06.07.495232PMC PMC10777399

Preprint — published as 10.1021/acs.jcim.3c01381

Open this paper in CoChat — free accountOpen PDF

Abstract

The rapid development of deep learning-based methods has considerably advanced the field of protein structure prediction. The accuracy of predicting the 3D structures of simple proteins is comparable to that of experimentally determined structures, providing broad possibilities for structure-based biological studies. Another critical question is whether and how multistate structures can be predicted from a given protein sequence. In this study, analysis of multiple two-state proteins demonstrated that deep learning-based contact map predictions contain structural information on both states, which suggests that it is probably appropriate to change the target of deep learningbased protein structure prediction from one specific structure to multiple likely structures. Furthermore, by combining deep learning- and physics-based computational methods, we developed a protocol for exploring alternative conformations from a known structure of a given protein, by which we successfully approached the holo-state conformation of a leucine-binding protein from its apo-state structure.

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 · 10.1101/2022.06.07.495232

Cites

2 of 2 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1 Analysis of contacts in the PCMs for the additional 91 proteins. The number of apo-/holo-state-specific contacts (“101”/“011” in Table S1) were represented on the x/y axis. The size of the circles represented the proteins’ sequence length. Only medium- and long-range contacts are considered. The PCMs of 84 proteins cover certain amounts of the state-specific contacts of both states.
Figure 1 Analysis of contacts in the PCMs for the additional 91 proteins. The number of apo-/holo-state-specific contacts (“101”/“011” in Table S1) were represented on the x/y axis. The size of the circles represented the proteins’ sequence length. Only medium- and long-range contacts are considered. The PCMs of 84 proteins cover certain amounts of the state-specific contacts of both states.CC BY-NC-ND · PubMed Central
Figure 2 Workflow of our prediction method. We started with a known protein state structure and used the information extracted from its TCM and PCM to explore the alternative state structure.
Figure 2 Workflow of our prediction method. We started with a known protein state structure and used the information extracted from its TCM and PCM to explore the alternative state structure.CC BY-NC-ND · PubMed Central
Figure 3 Precision in the DCM prediction for the 91 two-state proteins. X-axis and y-axis represent the prediction precision before and after considering the flexibility of known protein structures, respectively. The precision was calculated as the number of true positives/number of all predicted positives.
Figure 3 Precision in the DCM prediction for the 91 two-state proteins. X-axis and y-axis represent the prediction precision before and after considering the flexibility of known protein structures, respectively. The precision was calculated as the number of true positives/number of all predicted positives.CC BY-NC-ND · PubMed Central
Figure 4 Extracting the likely contacts in the alternative conformation from the CM analysis, using LBP as an example. (A) Apo (blue, PDB ID: 1usg) and holo (gray, PDB ID: 1usi) structures of LBP showed a clear open–closed transition. (B) Comparison between the contact map from the holo state (TCM1, lower left) and the predicted contact map from its sequence (PCM, upper right) shows additional contacts of high prediction confidence (gray circle), which may represent new contacts in the alternative conformation. The star points are in correspondence with residue pairs in Table 3, where red stars represent residue pairs in bold and blue ones represent the rest. (C) The DCM is also shown for visualization. (D) Distances (in Å) between the four selected pairs of residues in the original apo-state structure, which were predicted to form contacts in the alternative state conformation.
Figure 4 Extracting the likely contacts in the alternative conformation from the CM analysis, using LBP as an example. (A) Apo (blue, PDB ID: 1usg) and holo (gray, PDB ID: 1usi) structures of LBP showed a clear open–closed transition. (B) Comparison between the contact map from the holo state (TCM1, lower left) and the predicted contact map from its sequence (PCM, upper right) shows additional contacts of high prediction confidence (gray circle), which may represent new contacts in the alternative conformation. The star points are in correspondence with residue pairs in Table 3, where red stars represent residue pairs in bold and blue ones represent the rest. (C) The DCM is also shown for visualization. (D) Distances (in Å) between the four selected pairs of residues in the original apo-state structure, which were predicted to form contacts in the alternative state conformation.CC BY-NC-ND · PubMed Central
Figure 5 Predicted contacts in the alternative state for the nine representative proteins. The residue pairs to form new contacts are linked with purple lines on the initial apo-state structures. Black arrows indicate the direction of conformational changes by inspecting the structures of both states.
Figure 5 Predicted contacts in the alternative state for the nine representative proteins. The residue pairs to form new contacts are linked with purple lines on the initial apo-state structures. Black arrows indicate the direction of conformational changes by inspecting the structures of both states.CC BY-NC-ND · PubMed Central
Figure 6 Modeling the alternative conformation of LBP with Rosetta. (A) Restraint energy function for the selected residue pairs during the optimization procedure. The initial distances between the four selected residue pairs are shown with cyan circles. (B) Scoring of the modeled structures and their RMSD with respect to the holo structure. 100 models were generated using Rosetta, and the RMSD between the apo and holo (target) states was indicated by dotted lines. The top 10 models (with the lowest scores) are represented in blue. The contact restraints were applied only on the four selected residue pairs. The correlation coefficient r between RMSD and energy score is labeled, too. (C) Best predicted conformation (yellow), overlapping with the closed-state structure (gray). The final distances between the selected residue pairs are shown in the enlarged view. (D) Similar to (B), but all 78 new contacts in the DCM were employed and restrained during the modeling.
Figure 6 Modeling the alternative conformation of LBP with Rosetta. (A) Restraint energy function for the selected residue pairs during the optimization procedure. The initial distances between the four selected residue pairs are shown with cyan circles. (B) Scoring of the modeled structures and their RMSD with respect to the holo structure. 100 models were generated using Rosetta, and the RMSD between the apo and holo (target) states was indicated by dotted lines. The top 10 models (with the lowest scores) are represented in blue. The contact restraints were applied only on the four selected residue pairs. The correlation coefficient r between RMSD and energy score is labeled, too. (C) Best predicted conformation (yellow), overlapping with the closed-state structure (gray). The final distances between the selected residue pairs are shown in the enlarged view. (D) Similar to (B), but all 78 new contacts in the DCM were employed and restrained during the modeling.CC BY-NC-ND · PubMed Central
Figure 7 Assessment of the modeled structures and their RMSD with respect to the target structures. Similar to Figure 6B, 100 models were generated for each protein with only the selected residue pairs subject to contact restraints. The RMSD between the apo and holo (target) states are indicated by dotted lines. The top 10 models (with the lowest scores) are represented in blue, while the worst 10 models (with the highest scores) were discarded. The remaining models are depicted in green. The fraction of the blue circles that are located to the left of the dashed lines is indicated in the upper left corner of each panel.
Figure 7 Assessment of the modeled structures and their RMSD with respect to the target structures. Similar to Figure 6B, 100 models were generated for each protein with only the selected residue pairs subject to contact restraints. The RMSD between the apo and holo (target) states are indicated by dotted lines. The top 10 models (with the lowest scores) are represented in blue, while the worst 10 models (with the highest scores) were discarded. The remaining models are depicted in green. The fraction of the blue circles that are located to the left of the dashed lines is indicated in the upper left corner of each panel.CC BY-NC-ND · PubMed Central
Figure 8 Comparison of the initial (blue), target (gray), and the best-modeled structures (yellow). The best-modeled structures represent the model with the lowest RMSD with respect to the target structure.
Figure 8 Comparison of the initial (blue), target (gray), and the best-modeled structures (yellow). The best-modeled structures represent the model with the lowest RMSD with respect to the target structure.CC BY-NC-ND · PubMed Central
Figure 9 Modeling the active conformation of rhodopsin from the predicted inactive structure by AlphaFold2. (A) The overlaid inactive crystal structure (blue, PDB ID: 2i35) and the predicted inactive structure by AlphaFold2 (pink). The RMSD between the two structures is 1.19 Å. (B) The predicted seven new contacts in the alternative state are indicated with purple lines on the AlphaFold2-predicted structure. (C) The best-sampled model in our prediction (yellow) successfully approached the active structure (gray, PDB ID: 6oy9) of rhodopsin, with an RMSD of 1.48 Å for helices 5 and 6. (D) Assessment of the 100 active-state models predicted using Rosetta and our protocol. The dashed line at 2.56 Å represents the RMSD between the AlphaFold2-predicted inactive model and active-state crystal structure (for helices 5 and 6).
Figure 9 Modeling the active conformation of rhodopsin from the predicted inactive structure by AlphaFold2. (A) The overlaid inactive crystal structure (blue, PDB ID: 2i35) and the predicted inactive structure by AlphaFold2 (pink). The RMSD between the two structures is 1.19 Å. (B) The predicted seven new contacts in the alternative state are indicated with purple lines on the AlphaFold2-predicted structure. (C) The best-sampled model in our prediction (yellow) successfully approached the active structure (gray, PDB ID: 6oy9) of rhodopsin, with an RMSD of 1.48 Å for helices 5 and 6. (D) Assessment of the 100 active-state models predicted using Rosetta and our protocol. The dashed line at 2.56 Å represents the RMSD between the AlphaFold2-predicted inactive model and active-state crystal structure (for helices 5 and 6).CC BY-NC-ND · 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: 60

  1. 1995, The protein folding problem and tertiary structure prediction, FEBS Letters, 361, 129, 10.1016/0014-5793(99)90003-0 · DOI
  2. 10.1126/science.1219021 · DOI · Paper page
  3. 10.1038/nbt.2419 · DOI · Paper page
  4. 2017, Accurate de novo prediction of protein contact map by ultra-deep learning model, PLoS Computational Biology, 13, e1005324, 10.1371/journal.pcbi.1005324 · DOI · Paper page
  5. 2019, A further leap of improvement in tertiary structure prediction in CASP13 prompts new routes for future assessments, Proteins: Structure, Function, and Bioinformatics, 87, 1100, 10.1002/prot.25787 · DOI · Paper page
  6. 2019, Deep-learning contactmap guided protein structure prediction in CASP13, Proteins: Structure, Function, and Bioinformatics, 87, 1149, 10.1002/prot.25792 · DOI · Paper page
  7. 2019, Analysis of distance-based protein structure prediction by deep learning in CASP13, Proteins: Structure, Function, and Bioinformatics, 87, 1069, 10.1002/prot.25810 · DOI
  8. 10.1126/science.abj8754 · DOI · Paper page
  9. 10.1016/j.sbi.2021.01.007 · DOI · Paper page
  10. 2021, Highly accurate protein structure prediction for the human proteome, Nature, 596, 590, 10.1038/s41586-021-03828-1 · DOI · Paper page
  11. 2021, Assessment of the CASP14 assembly predictions, Proteins, 89, 1787, 10.1002/prot.26199 · DOI · Paper page
  12. 10.1038/nmeth.3213 · DOI · Paper page
  13. 10.1371/journal.pcbi.1006805 · DOI
  14. 2020, High-accuracy protein structures by combining machine-learning with physics-based refinement, Proteins: Structure, Function and Bioinformatics, 88, 637, 10.1002/prot.25847 · DOI
  15. 2021, Physics-based protein structure refinement in the era of artificial intelligence, Proteins, 89, 1870, 10.1002/prot.26161 · DOI
  16. 2020, Discovering protein conformational flexibility through artificial-intelligence-aided molecular dynamics, Journal of Physical Chemistry B, 124, 8221, 10.1021/acs.jpcb.0c03985 · DOI
  17. 10.1002/prot.21510 · DOI
  18. 10.1021/bi0260839 · DOI
  19. 10.1038/nature11896 · DOI
  20. 10.1016/j.neuron.2015.08.032 · DOI · Paper page
  21. 10.1016/j.mam.2012.07.001 · DOI · Paper page
  22. Heo, L. ; Feig, M. Multi-state modeling of G-protein coupled receptors at experimental accuracy. Proteins: Structure, Function, and Bioinformatics 2022, · DOI · Paper page
  23. 10.1038/nature06522 · DOI · Paper page
  24. 10.1016/j.cell.2015.03.049 · DOI
  25. 10.1126/science.1208351 · DOI · Paper page
  26. Orellana, L. ; Yoluk, O. ; Carrillo, O. ; Orozco, M. ; Lindahl, E. Prediction and validation of protein intermediate states from structurally rich ensembles and coarse-grained simulations. Nature Communications 2016, 7. · DOI
  27. 2019, Biophysical techniques in structural biology, Annual Review of Biochemistry, 88, 25, 10.1146/annurev-biochem-013118-111947 · DOI
  28. 2019, Protein Data Bank: The single global archive for 3D macromolecular structure data, Nucleic Acids Research, 47, D520, 10.1093/nar/gky949 · DOI · Paper page
  29. 10.1073/pnas.1315625110 · DOI
  30. 10.1016/j.str.2015.10.025 · DOI
  31. 2021, Probing allosteric regulations with coevolution-driven molecular simulations, Science Advances, 7, eabj0786, 10.1126/sciadv.abj0786 · DOI
  32. 2022, Uncovering the Dominant Motion Modes of Allosteric Regulation Improves Allosteric Site Prediction, Journal of Chemical Information and Modeling, 62, 187, 10.1021/acs.jcim.1c01267 · DOI
  33. 10.1093/bioinformatics/btv390 · DOI
  34. 10.1073/pnas.1821309116 · DOI · Paper page
  35. 2020, Difference contact maps: From what to why in the analysis of the conformational flexibility of proteins, PLoS ONE, 15, e0226702, 10.1371/journal.pone.0226702 · DOI
  36. 2020, FingerprintContacts: Predicting alternative conformations of proteins from coevolution, Journal of Physical Chemistry B, 124, 3605, 10.1021/acs.jpcb.9b11869 · DOI
  37. 10.1103/PhysRevLett.82.656 · DOI
  38. 10.1016/j.molcel.2019.06.007 · DOI
  39. 10.1016/j.molcel.2012.09.023 · DOI
  40. 10.1074/jbc.M311890200 · DOI
  41. Saldaño, T. et al. Impact of protein conformational diversity on AlphaFold predictions. Bioinformatics 2022, btac202. · DOI · Paper page
  42. 2015, Protein structure determination by combining sparse NMR data with evolutionary couplings, Nature Methods, 12, 751, 10.1038/nmeth.3455 · DOI
  43. 10.1073/pnas.1914677117 · DOI · Paper page
  44. 2021, Predicting new protein conformations from molecular dynamics simulation conformational landscapes and machine learning, Proteins: Structure, Function, and Bioinformatics, 89, 915, 10.1002/prot.26068 · DOI
  45. 2022, Membrane contact probability: An essential and predictive character for the structural and functional studies of membrane proteins, PLoS Computational Biology, 18, e1009972, 10.1371/journal.pcbi.1009972 · DOI · Paper page
  46. 10.1021/cr4003769 · DOI
  47. 10.1016/j.molcel.2007.08.022 · DOI
  48. 2019, The Hsp70 chaperone network, Nature Reviews Molecular Cell Biology, 20, 665, 10.1038/s41580-019-0133-3 · DOI · Paper page
  49. 10.1111/j.1365-2958.1996.tb02484.x · DOI
  50. 2012, HHblits: Lightning-fast iterative protein sequence searching by HMM-HMM alignment, Nature Methods, 9, 173, 10.1038/nmeth.1818 · DOI · Paper page
  51. Sikic, K. ; Jeren, B. ; Tomic, S. The protein secondary structure flexibility. Proceedings of the 2008 International Conference on Bioinformatics and Computational Biology 2008, 918–926.
  52. 10.1093/bioinformatics/btq007 · DOI
  53. 10.1073/pnas.96.10.5482 · DOI
  54. 2016, Simultaneous optimization of biomolecular energy functions on features from small molecules and macromolecules, Journal of Chemical Theory and Computation, 12, 6201, 10.1021/acs.jctc.6b00819 · DOI
  55. 10.1016/J.SOFTX.2015.06.001 · DOI · Paper page
  56. 2016, CHARMM36m: An improved force field for folded and intrinsically disordered proteins, Nature Methods, 14, 71
  57. 10.1063/1.448118 · DOI
  58. 10.1063/1.464397 · DOI
  59. 2018, Faster simulations with a 5 fs time step for lipids in the CHARMM force field, Journal of Chemical Theory and Computation, 14, 3342, 10.1021/acs.jctc.8b00267 · DOI
  60. 10.1016/0263-7855(96)00018-5 · DOI · Paper page

Metadata

Authors
Li, J., Wang, L., Zhu, Z., & Song, C.
Affiliations
Peking University (CN)
Corresponding author
Song, Chen
ORCID
Venue
Journal of Chemical Information and Modeling
Publisher
openRxiv
Published
2022-06-09
Volume
64
Issue
1
Pages
301-315
ISSN
1549-9596 (ISSN-L)
DOI
10.1101/2022.06.07.495232
PMC
PMC10777399
OpenAlex
W4389993920
Cited by
21
References
2
Citations by year
2026: 5 · 2025: 10 · 2024: 4 · 2023: 1 · 2022: 1
FWCI
1.78
Type
journal article
Language
en
Open access
Licence
cc-by-nc-nd
Locations
  • Journal of Chemical Information and Modeling (publishedVersion, cc-by-nc-nd)
  • bioRxiv (Cold Spring Harbor Laboratory) (acceptedVersion, cc-by-nc-nd)
  • PubMed Central (publishedVersion, cc-by-nc-nd)
  • PubMed (publishedVersion)
Funders
  • National Natural Science Foundation of China (21873006, 32071251)
  • Peking University
  • National Key Research and Development Program of China (2021YFE0108100)
Topics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Protein Structure and Dynamics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Machine Learning in Bioinformatics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Structural Biology › Enzyme Structure and Function
Keywords
contact map prediction, two-state proteins, deep learning protein structure prediction, conformational change
MeSH terms
Amino Acid Sequence, Protein Conformation, Proteins, Computational Biology

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