CoChat
Sign Up
PreprintCited by 12SARS-CoV-2 and COVID-19 ResearchProtein Structure and DynamicsComputational Drug Discovery Methods

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

AI-driven prediction of SARS-CoV-2 variant binding trends from atomistic simulations

Open access
Capponi, Sara, Wang, Shangying, Navarro, Erik J., Bianco, Simone·2021·The European Physical Journal E
DOI 10.1101/2021.03.07.434295PMC PMC8493367

Preprint — published as 10.1140/epje/s10189-021-00119-5 · Paper page

Open this paper in CoChat — free accountOpen PDF

Abstract

We present a novel technique to predict binding affinity trends between two molecules from atomistic molecular dynamics simulations. The technique uses a neural network algorithm applied to a series of images encoding the distance between two molecules in time. We demonstrate that our algorithm is capable of separating with high accuracy non-hydrophobic mutations with low binding affinity from those with high binding affinity. Moreover, we show high accuracy in prediction using a small subset of the simulation, therefore requiring a much shorter simulation time. We apply our algorithm to the binding between several variants of the SARS-CoV-2 spike protein and the human receptor ACE2.

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 · 10.1101/2021.03.07.434295

Cites

1 of 1 references verified0 citation errors · 0 not checkable

Statistics

Figures

Fig. 1 a Representative snapshot of the chimeric spike receptor binding domain (RBD) (purple) in association with ACE2 (gray). Proteins are represented in New Cartoon format. Zinc (orange), and chloride (blue) ions are represented in van der Waals format. Bottom left: representation of the spike trimeric complex from [9], one chain is blue with the RBD colored purple, the other 2 chains are transparent gray and green for clarity. b, c Closeup view of the contact region between RBD and ACE2. The non-hydrophobic, non-aromatic residues that are mutated and studied in this paper are indicated. They are represented in licorice format colored according to the ion type (red for oxygen, cyan for carbon, blue for nitrogen). Hydrogen atoms are not displayed for clarity. c Same representation of panel b rotated of 90^(∘)
Fig. 1 a Representative snapshot of the chimeric spike receptor binding domain (RBD) (purple) in association with ACE2 (gray). Proteins are represented in New Cartoon format. Zinc (orange), and chloride (blue) ions are represented in van der Waals format. Bottom left: representation of the spike trimeric complex from [9], one chain is blue with the RBD colored purple, the other 2 chains are transparent gray and green for clarity. b, c Closeup view of the contact region between RBD and ACE2. The non-hydrophobic, non-aromatic residues that are mutated and studied in this paper are indicated. They are represented in licorice format colored according to the ion type (red for oxygen, cyan for carbon, blue for nitrogen). Hydrogen atoms are not displayed for clarity. c Same representation of panel b rotated of 90^(∘)CC BY · PubMed Central
Fig. 2 Conversion of the MD data to input matrices for a neural algorithm. a Representative snapshot of the chimeric RBD-ACE2 complex indicating the Cα carbon atoms used to calculate the distance between the RBD and ACE2. The color code is as in Fig. 1, and the Cα carbon atoms are represented in cyan, van der Waals format. The distances between the atoms are then represented as matrix, are normalized by each of their minima and maxima, and converted to gray-scaled images. b Representative distance matrices of the chimeric RBD-ACE2 complex simulations taken at different steps of the simulation showing the effect of data preprocessing. c Representation of the data processing to obtain machine learning inputs. We first subtract the distance matrix at t=0 from the distance matrices from the simulations at each time steps (central panel). We then use a moving average to average the subtracted matrices and use these as input to the algorithm (right-hand panel)
Fig. 2 Conversion of the MD data to input matrices for a neural algorithm. a Representative snapshot of the chimeric RBD-ACE2 complex indicating the Cα carbon atoms used to calculate the distance between the RBD and ACE2. The color code is as in Fig. 1, and the Cα carbon atoms are represented in cyan, van der Waals format. The distances between the atoms are then represented as matrix, are normalized by each of their minima and maxima, and converted to gray-scaled images. b Representative distance matrices of the chimeric RBD-ACE2 complex simulations taken at different steps of the simulation showing the effect of data preprocessing. c Representation of the data processing to obtain machine learning inputs. We first subtract the distance matrix at t=0 from the distance matrices from the simulations at each time steps (central panel). We then use a moving average to average the subtracted matrices and use these as input to the algorithm (right-hand panel)CC BY · PubMed Central
Fig. 3 Sample gray-scaled images representing the averaged matrices for some of the mutations studied in this paper at t=20ns
Fig. 3 Sample gray-scaled images representing the averaged matrices for some of the mutations studied in this paper at t=20nsCC BY · PubMed Central
Fig. 4 Structure of the convolutional neural network and ensemble prediction used in this study. a The convolutional neural network architecture. b The ensemble prediction consisting of N independently trained convolutional neural networks. The final prediction is based on the simple majority of all neural networks. c Definition of the output categories
Fig. 4 Structure of the convolutional neural network and ensemble prediction used in this study. a The convolutional neural network architecture. b The ensemble prediction consisting of N independently trained convolutional neural networks. The final prediction is based on the simple majority of all neural networks. c Definition of the output categoriesCC BY · PubMed Central
Fig. 5 The overall test accuracy of CNNs varies with the simulation time. The blue dots are the average prediction accuracy from ten independently trained CNNs. The error bars represent the 10th percentile and 80th percentile. The orange diamonds represent the prediction accuracy of the ensemble prediction, voted by 10 NN assembles
Fig. 5 The overall test accuracy of CNNs varies with the simulation time. The blue dots are the average prediction accuracy from ten independently trained CNNs. The error bars represent the 10th percentile and 80th percentile. The orange diamonds represent the prediction accuracy of the ensemble prediction, voted by 10 NN assemblesCC BY · PubMed Central
Fig. 6 RMSD calculated for the Cα carbon atoms of the RBD and ACE2 and for the contact area for the simulations reported in the panels. For clarity, data are represented every 2 ns
Fig. 6 RMSD calculated for the Cα carbon atoms of the RBD and ACE2 and for the contact area for the simulations reported in the panels. For clarity, data are represented every 2 nsCC BY · PubMed Central
Fig. 7 RMSD calculated for the Cα carbon atoms of the RBD and ACE2 and for the contact area for the simulations reported in the panels. For clarity, data are represented every 2 ns
Fig. 7 RMSD calculated for the Cα carbon atoms of the RBD and ACE2 and for the contact area for the simulations reported in the panels. For clarity, data are represented every 2 nsCC BY · PubMed Central
Fig. 8 RMSD calculated for the Cα carbon atoms of the RBD and ACE2 and for the contact area for the simulations reported in the panels. For clarity, data are represented every 2 ns
Fig. 8 RMSD calculated for the Cα carbon atoms of the RBD and ACE2 and for the contact area for the simulations reported in the panels. For clarity, data are represented every 2 nsCC 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: 67

  1. H. J. Maier , E. Bickerton , and P. Britton , Coronaviruses: Methods and Protocols, 1–282 (2015).https://doi.org/10.1007/978-1-4939-2438-7 · DOI
  2. 2016, Annu. Rev. Virol, 3, 237, 10.1146/annurev-virology-110615-042301 · DOI · Paper page
  3. 10.3390/v4061011 · DOI · Paper page
  4. 2005, Virology, 335, 176
  5. 10.1073/pnas.1708727114 · DOI
  6. 10.1128/JVI.77.16.8801-8811.2003 · DOI · Paper page
  7. 2019, Advances in Virus Research, 105, 94
  8. 10.1126/science.abb2507 · DOI · Paper page
  9. 10.1016/j.cell.2020.02.035 · DOI
  10. 2020, Nature, 581, 215, 10.1038/s41586-020-2180-5 · DOI · Paper page
  11. 10.1038/s41586-020-2179-y · DOI · Paper page
  12. 10.1016/j.cell.2020.05.034 · DOI
  13. 10.1126/science.abb2762 · DOI
  14. 10.1038/s41586-020-2012-7 · DOI · Paper page
  15. 2020, Cell, 182, 1295, 10.1016/j.cell.2020.08.012 · DOI · Paper page
  16. 2021, Cell Host and Microbe, 29, 44, 10.1016/j.chom.2020.11.007 · DOI · Paper page
  17. 2020, Pharmacology and Translational Science, 3, 1023, 10.1021/acsptsci.0c00113 · DOI
  18. 2020, ACS Central Science, 6, 1654, 10.1021/acscentsci.0c01236 · DOI
  19. C. Peng , et al, ChemrXiv (2020).https://doi.org/10.26434/chemrxiv.11877492.v1 · DOI
  20. 2020, Journal of Computer-Aided Molecular Design, 35, 195
  21. 2021, J. Phys. Chem. Lett, 21, 5494
  22. 10.1073/pnas.2008209117 · DOI
  23. 10.1021/acs.jmedchem.1c00311 · DOI
  24. 2020, Sci. R, 10, 14214
  25. 2021, Biophys. J, 120, 1072, 10.1016/j.bpj.2020.10.036 · DOI
  26. 2020, J. Phys. Chem. B, 140, 10034
  27. 2020, J. Chem. Inf. Model, 60, 5794, 10.1021/acs.jcim.0c00679 · DOI
  28. J. M. Delgado and N. Duro and D. M. Rogers and A. Tkatchenko and S. A. Pandit and S. Varma , Proteins, 1–11, (2021)https://doi.org/10.1002/prot.260865794-5802 · DOI
  29. 2020, J. Phys. Chem. B, 124, 7336, 10.1021/acs.jpcb.0c04511 · DOI
  30. 10.1038/nchem.2785 · DOI
  31. G. R. Bowman , V. S. Pande , F. Noé . Dordrecht: Springer Science and Business Media. 148(2014).
  32. 10.1063/1.1472510 · DOI
  33. 10.1016/S0006-3495(96)79552-8 · DOI
  34. 2017, Annu Rev Biophys, 46, 43, 10.1146/annurev-biophys-070816-033834 · DOI
  35. 2011, J. Comput. Chem, 13, 2810
  36. 2009, Biophys J, 7, 2004
  37. C. Chipot and A. Pohorille , Free Energy Calculations, Springer (2007). · DOI
  38. 2020, J. Phys. Chem. Lett, 11, 9656, 10.1021/acs.jpclett.0c03130 · DOI
  39. 2020, Lecture Notes in Physics, 968, 331, 10.1007/978-3-030-40245-7_16 · DOI
  40. L. Casalino , et al, BioRxiv (2020).https://doi.org/10.1101/2020.11.19.390187 · DOI
  41. 2019, Massive computational acceleration by using neural networks to emulate mechanismbased biological models, Nature Communications, 10, 19
  42. 2020, Predictive biology: modelling, understanding and harnessing microbial complexity, Nature Reviews Microbiology, 18, 507, 10.1038/s41579-020-0372-5 · DOI
  43. 2020, J. Phys. Chem. b, 124, 9103, 10.1021/acs.jpcb.0c05977 · DOI
  44. 2020, Journal of Chemical Information and Modeling, 60, 2766, 10.1021/acs.jcim.0c00026 · DOI
  45. 10.1016/j.bpj.2019.12.016 · DOI · Paper page
  46. 2019, Bioinformatics, 35, i269, 10.1093/bioinformatics/btz339 · DOI
  47. 10.1038/s41586-020-2179-y · DOI · Paper page
  48. 2016, J. Chem. Thory Comput, 12, 405, 10.1021/acs.jctc.5b00935 · DOI
  49. 2008, J. Chem. Thory Comput, 29, 1859, 10.1002/jcc.20945 · DOI · Paper page
  50. 10.1016/0263-7855(96)00018-5 · DOI · Paper page
  51. 10.1002/jcc.20289 · DOI · Paper page
  52. 2005, J. Chem. Phys, 153, 044130
  53. 10.1038/nmeth.4067 · DOI
  54. 10.1063/1.445869 · DOI
  55. 1993, J. Chem. Phys, 103, 4613
  56. 10.1063/1.467468 · DOI
  57. 10.1063/1.464397 · DOI
  58. 10.1063/1.470117 · DOI
  59. 10.1080/08927029108022142 · DOI
  60. 10.1016/0021-9991(77)90098-5 · DOI
  61. R. J. Gowers , et al. In S. Benthall and S. Rostrup , editors,Proceedings of the 15th Python in Science Conference, 98–105 (2016).
  62. 10.1002/jcc.21787 · DOI
  63. 2021, Collective colony growth is optimized by branching pattern formation in Pseudomonas aeruginosa, Molecular Systems Biology, 17, e10089, 10.15252/msb.202010089 · DOI
  64. G. Schreiber and G. Haran and H.-X. Zhou , Chem. Rev., 109, (2008). · DOI
  65. 10.1073/pnas.2133366100 · DOI
  66. 2019, Biochemistry, 58, 156, 10.1021/acs.biochem.8b00977 · DOI
  67. 2020, Nature Machine Intelligence, 2, 116, 10.1038/s42256-020-0149-6 · DOI · Paper page

Metadata

Authors
Capponi, S., Wang, S., Navarro, E. J., & Bianco, S.
Affiliations
University of California, San Francisco (US)
Corresponding author
Bianco, Simone
ORCID
Venue
The European Physical Journal E
Publisher
openRxiv
Published
2021-03-08
Volume
44
Issue
10
Pages
123-123
ISSN
1292-8941 (ISSN-L)
DOI
10.1101/2021.03.07.434295
PMC
PMC8493367
OpenAlex
W3202379675
Cited by
12
References
1
Citations by year
2026: 2 · 2025: 2 · 2024: 3 · 2023: 1 · 2022: 3 · 2021: 1
FWCI
0.63
Type
journal article
Language
en
Open access
Licence
Locations
  • The European Physical Journal E (publishedVersion, cc-by)
  • bioRxiv (Cold Spring Harbor Laboratory) (acceptedVersion, cc-by-nc)
  • Europe PMC (PubMed Central) (publishedVersion, other-oa)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • Springer Link (Chiba Institute of Technology) (submittedVersion)
Funders
  • National Science Foundation (1548297)
  • Rensselaer Polytechnic Institute
Topics
  • Health Sciences › Medicine › Infectious Diseases › SARS-CoV-2 and COVID-19 Research
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Protein Structure and Dynamics
  • Life Sciences › Pharmacology, Toxicology and Pharmaceutics › Drug Discovery › Computational Drug Discovery Methods
Keywords
SARS-CoV-2 variants, spike protein, ACE2 receptor, protein-protein binding affinity, binding affinity prediction, artificial neural networks, molecular dynamics simulation, atomistic simulation
MeSH terms
Angiotensin-Converting Enzyme 2, SARS-CoV-2, Algorithms, Artificial Intelligence, Humans, Models, Molecular, Mutation, Protein Binding, Protein Conformation, Neural Networks, Computer, Molecular Dynamics Simulation, Spike Glycoprotein, Coronavirus

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