Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

  • Article
  • Published:

Extraction of protein dynamics information from cryo-EM maps using deep learning

Nature Machine Intelligence volume 3, pages 153–160 (2021)Cite this article

Subjects

A preprint version of the article is available at bioRxiv.

Abstract

Elucidation of both the three-dimensional structure and the dynamics of a protein is essential to understand its function. Technical breakthroughs in single-particle analysis based on cryo-electron microscopy (cryo-EM) have enabled the three-dimensional structures of numerous proteins to be solved at atomic or near-atomic resolution. However, the analysis of the dynamics of protein targets using cryo-EM is often challenging because of their large sizes and complex structural assemblies. Here, we describe DEFMap, a deep learning-based approach to directly extract the dynamics associated with the atomic fluctuations that are hidden in cryo-EM density maps. Using only cryo-EM density data, DEFMap provides dynamics that correlate well with data obtained from molecular dynamics simulations and experimental approaches. Furthermore, DEFMap successfully detects changes in dynamics that are associated with molecular recognition. This strategy combines deep learning, experimental data and molecular dynamics simulations, and may reveal a new multidisciplinary approach for protein science.

This is a preview of subscription content, access via your institution

Access options

Buy this article

  • Purchase on SpringerLink
  • Instant access to the full article PDF.

USD 39.95

Prices may be subject to local taxes which are calculated during checkout

Fig. 1: DEFMap-based extraction of dynamics features from cryo-EM maps.
Fig. 2: DEFMap performance for three proteins not included in the training dataset.
Fig. 3: DEFMap-based detection of dynamics changes induced by ligand binding.
Fig. 4: Visual representation of DEFMap-derived dynamics for the spike protein in SARS-CoV-2 and a ZIKV particle.

Similar content being viewed by others

Data availability

All of the datasets used in this study are publicly available in the Protein Data Bank (PDB) at https://www.rcsb.org/ and the Electron Microscopy Data Bank (EMDB) at https://www.ebi.ac.uk/pdbe/emdb/. Detailed descriptions of the datasets are provided in Supplementary Tables 1,2,3,4. The models and the preprocessed input data used in 25-fold cross-validation are available in the Zenodo public repository at https://doi.org/10.5281/zenodo.4317158.

Code availability

The Python code used to implement DEFMap is available on GitHub at https://github.com/clinfo/DEFMap (https://doi.org/10.5281/zenodo.4317158). DEFMap is available under the MIT License.

References

  1. Boehr, D. D., Nussinov, R. & Wright, P. E. The role of dynamic conformational ensembles in biomolecular recognition. Nat. Chem. Biol. 5, 789–796 (2009).

    Article  Google Scholar 

  2. Kohen, A. Role of dynamics in enzyme catalysis: substantial versus semantic controversies. Acc. Chem. Res. 48, 466–473 (2015).

    Article  Google Scholar 

  3. Cheng, Y. Single-particle cryo-EM—how did it get here and where will it go. Science 361, 876–880 (2018).

    Article  Google Scholar 

  4. Cheng, Y., Grigorieff, N., Penczek, P. A. & Walz, T. A primer to single-particle cryo-electron microscopy. Cell 161, 438–449 (2015).

    Article  Google Scholar 

  5. Murata, K. & Wolf, M. Cryo-electron microscopy for structural analysis of dynamic biological macromolecules. Biochim. Biophys. Acta Gen. Subj. 1862, 324–334 (2018).

    Article  Google Scholar 

  6. Nitta, R., Imasaki, T. & Nitta, E. Recent progress in structural biology: lessons from our research history. Microscopy 67, 187–195 (2018).

    Article  Google Scholar 

  7. Masson, G. R. et al. Recommendations for performing, interpreting and reporting hydrogen deuterium exchange mass spectrometry (HDX-MS) experiments. Nat. Methods 16, 595–602 (2019).

    Article  Google Scholar 

  8. Hollingsworth, S. A. & Dror, R. O. Molecular dynamics simulation for all. Neuron 99, 1129–1143 (2018).

    Article  Google Scholar 

  9. Kühlbrandt, W. Cryo-EM enters a new era. eLife 3, e03678 (2014).

    Article  Google Scholar 

  10. Cheng, Y. Single-particle cryo-EM at crystallographic resolution. Cell 161, 450–457 (2015).

    Article  Google Scholar 

  11. Merk, A. et al. Breaking cryo-EM resolution barriers to facilitate drug discovery. Cell 165, 1698–1707 (2016).

    Article  Google Scholar 

  12. Gremer, L. et al. Fibril structure of amyloid-β(1–42) by cryo-electron microscopy. Science 358, 116–119 (2017).

    Article  Google Scholar 

  13. Kato, T., Makino, F., Miyata, T., Horváth, P. & Namba, K. Structure of the native supercoiled flagellar hook as a universal joint. Nat. Commun. 10, 1–8 (2019).

    Article  Google Scholar 

  14. Kujirai, T. et al. Structural basis of the nucleosome transition during RNA polymerase II passage. Science 362, 595–598 (2018).

    Article  Google Scholar 

  15. Li, X. et al. A unified mechanism for intron and exon definition and back-splicing. Nature 573, 375–380 (2019).

    Article  Google Scholar 

  16. Vilas, J. L. et al. MonoRes: automatic and accurate estimation of local resolution for electron microscopy maps. Structure 26, 337–344.e4 (2018).

    Article  Google Scholar 

  17. Kucukelbir, A., Sigworth, F. J. & Tagare, H. D. Quantifying the local resolution of cryo-EM density maps. Nat. Methods 11, 63–65 (2014).

    Article  Google Scholar 

  18. Berman, H. M. et al. The protein data bank. Nucleic Acids Res. 28, 235–242 (2000).

    Article  Google Scholar 

  19. Hollingsworth, S. A. et al. Cryptic pocket formation underlies allosteric modulator selectivity at muscarinic GPCRs. Nat. Commun. 10, 3289 (2019).

    Article  Google Scholar 

  20. Plattner, N., Doerr, S., De Fabritiis, G. & Noé, F. Complete protein–protein association kinetics in atomic detail revealed by molecular dynamics simulations and Markov modelling. Nat. Chem. 9, 1005–1011 (2017).

    Article  Google Scholar 

  21. Ji, S., Xu, W., Yang, M. & Yu, K. 3D convolutional neural networks for human action recognition. IEEE Trans. Pattern Anal. Mach. Intell. 35, 221–231 (2013).

    Article  Google Scholar 

  22. Maturana, D. & Scherer, S. VoxNet: a 3D convolutional neural network for real-time object recognition. In Proc. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) 922–928 (IEEE, 2015).

  23. Zhirong, W. et al. 3D ShapeNets: a deep representation for volumetric shapes. In Proc. IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 1912–1920 (IEEE, 2015).

  24. Chen, H., Dou, Q., Yu, L., Qin, J. & Heng, P.-A. VoxResNet: deep voxelwise residual networks for brain segmentation from 3D MR images. NeuroImage 170, 446–455 (2018).

    Article  Google Scholar 

  25. Kamnitsas, K. et al. Efficient multi-scale 3D CNN with fully connected CRF for accurate brain lesion segmentation. Med. Image Anal. 36, 61–78 (2017).

    Article  Google Scholar 

  26. Qi, D. et al. Automatic detection of cerebral microbleeds from MR Images via 3D convolutional neural networks. IEEE Trans. Med. Imaging 35, 1182–1195 (2016).

    Article  Google Scholar 

  27. Avramov, T. K. et al. Deep learning for validating and estimating resolution of cryo-electron microscopy density maps. Molecules 24, 1181 (2019).

    Article  Google Scholar 

  28. Maddhuri Venkata Subramaniya, S. R., Terashi, G. & Kihara, D. Protein secondary structure detection in intermediate-resolution cryo-EM maps using deep learning. Nat. Methods 16, 911–917 (2019).

    Article  Google Scholar 

  29. Mostosi, P., Schindelin, H., Kollmannsberger, P. & Thorn, A. Automated interpretation of cryo-EM density maps with convolutional neural networks. bioRxiv, 644476 (2019).

  30. Xu, K., Wang, Z., Shi, J., Li, H. & Zhang, Q. C. A2-Net: molecular structure estimation from cryo-EM density volumes. In Proc. AAAI Conference on Artificial Intelligence 33, 1230–1237 (AAAI, 2019).

  31. Lawson, C. L. et al. EMDataBank.org: unified data resource for CryoEM. Nucleic Acids Res. 39, D456–D464 (2010).

    Article  Google Scholar 

  32. de la Rosa-Trevín, J. M. et al. Scipion: a software framework toward integration, reproducibility and validation in 3D electron microscopy. J. Struct. Biol. 195, 93–99 (2016).

    Article  Google Scholar 

  33. Goodfellow, I., Bengio, Y. & Courville, A. Deep Learning (MIT Press, 2016).

  34. Zhang, W., Lukoyanova, N., Miah, S., Lucas, J. & Vaughan, C. K. Insights into centromere DNA bending revealed by the Cryo-EM structure of the core centromere binding factor 3 with Ndc10. Cell Rep. 24, 744–754 (2018).

    Article  Google Scholar 

  35. Sun, Y. et al. Molecular basis for the recognition of the human AAUAAA polyadenylation signal. Proc. Natl Acad. Sci. USA 115, E1419–E1428 (2018).

    Google Scholar 

  36. Cash, J. N. et al. Cryo–electron microscopy structure and analysis of the P-Rex1–Gβγ signaling scaffold. Sci. Adv. 5, eaax8855 (2019).

    Article  Google Scholar 

  37. Dedden, D. et al. The architecture of Talin1 reveals an autoinhibition mechanism. Cell 179, 120–131.e13 (2019).

    Article  Google Scholar 

  38. Wongpalee, S. P. et al. CryoEM structures of Arabidopsis DDR complexes involved in RNA-directed DNA methylation. Nat. Commun. 10, 3916 (2019).

    Article  Google Scholar 

  39. Lan, P. et al. Structural insight into precursor tRNA processing by yeast ribonuclease P. Science 362, eaat6678 (2018).

    Article  Google Scholar 

  40. Abascal-Palacios, G., Ramsay, E. P., Beuron, F., Morris, E. & Vannini, A. Structural basis of RNA polymerase III transcription initiation. Nature 553, 301–306 (2018).

    Article  Google Scholar 

  41. Walls, A. C. et al. Structure, function, and antigenicity of the SARS-CoV-2 spike glycoprotein. Cell 181, 281–292.e6 (2020).

    Article  Google Scholar 

  42. Walls, A. C. et al. Tectonic conformational changes of a coronavirus spike glycoprotein promote membrane fusion. Proc. Natl Acad. Sci. USA 114, 11157–11162 (2017).

    Article  Google Scholar 

  43. Wrapp, D. et al. Cryo-EM structure of the 2019-nCoV spike in the prefusion conformation. Science 367, 1260–1263 (2020).

    Article  Google Scholar 

  44. Kostyuchenko, V. A. et al. Structure of the thermally stable Zika virus. Nature 533, 425–428 (2016).

    Article  Google Scholar 

  45. Wu, W. et al. Expression of quasi-equivalence and capsid dimorphism in the Hepadnaviridae. PLoS Comput. Biol. 16, e1007782 (2020).

    Article  Google Scholar 

  46. Cao, L. et al. Structural basis for neutralization of hepatitis A virus informs a rational design of highly potent inhibitors. PLoS Biol. 17, e3000229 (2019).

    Article  Google Scholar 

  47. Liu, Y. et al. Molecular basis for the acid-initiated uncoating of human enterovirus D68. Proc. Natl Acad. Sci. USA 115, E12209–E12217 (2018).

    Article  Google Scholar 

  48. Hamaguchi, T. et al. A new cryo-EM system for single particle analysis. J. Struct. Biol. 207, 40–48 (2019).

    Article  Google Scholar 

  49. Kato, T. et al. CryoTEM with a cold field emission gun that moves structural biology into a new stage. Microsc. Microanal. 25, 998–999 (2019).

    Article  Google Scholar 

  50. Ramírez-Aportela, E. et al. Automatic local resolution-based sharpening of cryo-EM maps. Bioinformatics 36, 765–772 (2019).

    Article  Google Scholar 

  51. Abraham, M. J. et al. GROMACS: high performance molecular simulations through multi-level parallelism from laptops to supercomputers. SoftwareX 1-2, 19–25 (2015).

    Article  Google Scholar 

  52. Lindorff-Larsen, K. et al. Improved side-chain torsion potentials for the Amber ff99SB protein force field. Proteins 78, 1950–1958 (2010).

    Article  Google Scholar 

  53. Jorgensen, W. L. & Thomas, L. L. Perspective on free-energy perturbation calculations for chemical equilibria. J. Chem. Theory Comput. 4, 869–876 (2008).

    Article  Google Scholar 

  54. Darden, T., York, D. & Pedersen, L. Particle mesh Ewald: an N⋅Log(N) method for Ewald sums in large systems. J. Chem. Phys. 98, 10089–10092 (1993).

    Article  Google Scholar 

  55. Hess, B. P-LINCS: a parallel linear constraint solver for molecular simulation. J. Chem. Theory Comput. 4, 116–122 (2008).

    Article  Google Scholar 

  56. Bussi, G., Donadio, D. & Parrinello, M. Canonical sampling through velocity rescaling. J. Chem. Phys. 126, 014101 (2007).

    Article  Google Scholar 

  57. Parrinello, M. & Rahman, A. Polymorphic transitions in single crystals: a new molecular dynamics method. J. Appl. Phys. 52, 7182–7190 (1981).

    Article  Google Scholar 

  58. Tang, G. et al. EMAN2: an extensible image processing suite for electron microscopy. J. Struct. Biol. 157, 38–46 (2007).

    Article  Google Scholar 

  59. Doerr, S., Harvey, M. J., Noé, F. & De Fabritiis, G. HTMD: high-throughput molecular dynamics for molecular discovery. J. Chem. Theory Comput. 12, 1845–1852 (2016).

    Article  Google Scholar 

  60. Chollet, F. Keras. https://keras.io (2015).

  61. Schrödinger LLC. The PyMOL Molecular Graphics System (version: 2.2.0) (2018).

  62. Pettersen, E. F. et al. UCSF Chimera—a visualization system for exploratory research and analysis. J. Comput. Chem. 25, 1605–1612 (2004).

    Article  Google Scholar 

Download references

Acknowledgements

This work was supported by MEXT as ‘Priority Issue on the Post K computer (Building Innovative Drug Discovery Infrastructure Through Functional Control of Biomolecular Systems)’ and as ‘Program for Promoting Researches on the Supercomputer Fugaku (Application of Molecular Dynamics Simulation to Precision Medicine Using Big Data Integration System for Drug Discovery)’. S.M. was supported by JSPS KAKENHI Grant Number JP17K15106.

Author information

Author notes

  1. These authors contributed equally: Shigeyuki Matsumoto, Shoichi Ishida.

Authors and Affiliations

  1. Medical Sciences Innovation Hub Program, RIKEN Cluster for Science, Technology and Innovation Hub, Tsurumi-ku, Kanagawa, Japan

    Shigeyuki Matsumoto, Kei Terayama & Yasushi Okuno

  2. Graduate School of Pharmaceutical Sciences, Kyoto University, Sakyo-ku, Kyoto, Japan

    Shoichi Ishida

  3. Graduate School of Medicine, Kyoto University, Sakyo-ku, Kyoto, Japan

    Mitsugu Araki, Kei Terayama & Yasushi Okuno

  4. Institute for Protein Research, Osaka University, Osaka, Japan

    Takayuki Kato

  5. Graduate School of Medical Life Science, Yokohama City University, Tsurumi-ku, Kanagawa, Japan

    Kei Terayama

  6. RIKEN Center for Advanced Intelligence Project, Tokyo, Japan

    Kei Terayama

Authors
  1. Shigeyuki Matsumoto
    View author publications

    Search author on:PubMed Google Scholar

  2. Shoichi Ishida
    View author publications

    Search author on:PubMed Google Scholar

  3. Mitsugu Araki
    View author publications

    Search author on:PubMed Google Scholar

  4. Takayuki Kato
    View author publications

    Search author on:PubMed Google Scholar

  5. Kei Terayama
    View author publications

    Search author on:PubMed Google Scholar

  6. Yasushi Okuno
    View author publications

    Search author on:PubMed Google Scholar

Contributions

S.M., S.I., K.T. and Y.O. designed the study. S.M. and S.I. performed the preprocessing of cryo-EM maps, designed the neural networks and carried out model training and DEFMap analyses. S.M. and M.A. performed the MD calculations. S.M. and T.K. prepared the original cryo-EM map data. K.T. and Y.O. conceived the project. S.M. and S.I. wrote the manuscript. All of the authors discussed the research, edited the manuscript and approved its final version.

Corresponding authors

Correspondence to Kei Terayama or Yasushi Okuno.

Ethics declarations

Competing interests

The authors declare no competing interests.

Additional information

Peer review information Nature Machine Intelligence thanks Joseph Davis, Carlos O. S. Sorzano and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Extended data

Extended Data Fig. 1 Correlation plot of the cryo-EM map local resolution (a) and the given density data (b) with DynamicsMD for EMD-3948 (entry 1 in Supplementary Table 2).

Normalized residue-averaged values are shown in the correlation plot. The regression lines are coloured orange; r denotes the correlation coefficient. The average correlation coefficient for the density data of 25 cryo-EM maps used in training is 0.459 ± 0.179 (Supplementary Table 2).

Extended Data Fig. 2 Correlation plots between MD-derived and DEFMap-determined dynamics for proteins not included in the training dataset.

a,b, Normalized atomic (a, blue) and residue-averaged (b, purple) values are shown in the correlation plots; the regression lines are coloured orange; r denotes the correlation coefficient.

Extended Data Fig. 3 The map resolution dependence of the dynamics extraction accuracy for test macromolecules (Supplementary Tables 3 and 4).

a,b, Correlation coefficient values between the MD- and DEFMap-derived atomic-specific (a) and residue-specific (b) dynamics at different resolutions. The cryo-EM maps of the training and the evaluation datasets were preprocessed to identical resolutions with low-pass filters.

Extended Data Fig. 4 DEFMap performance for a map with the intermediate resolution (EMDB/PDB entry, EMD-4772/6R9T).

a, Distribution of local resolution. The local resolution is calculated using the MonoRes implementation in Scipion and mapped onto the original map using different colours as indicated in the colour bars. b,c,d, The correlation plots between DynamicsDEFMap and DynamicsMD for the intermediate-resolution map. The target cryo-EM maps are processed by low-pass filters with cutoff values of 7 Å (b), 5 Å (c), and 6 Å (d). The predictions are carried out using the models trained by the datasets preprocessed to the identical resolutions. The regression lines are coloured orange; r denotes the correlation coefficient.

Extended Data Fig. 5 The spatial distribution of the fragment-specific dynamics in the atomic models (EMDB/PDB entry, EMD-20308/6PCV).

The representative fragments-specific dynamics for Rac exchanger 1 complexed with G protein beta gamma subunits are mapped onto the atomic model using different colours as indicated in the colour bar (HDX-MS, left; MD, middle; DEFMap, right). The fragments not detected in the HDX-MS experiments are coloured grey.

Extended Data Fig. 6 Residue-specific dynamics profiles for apo/holo proteins.

a,b, DynamicsMD (black) and DynamicsDEFMap (magenta) datasets for 6OIS/EMD-20080 and 6OIT/EMD-20081 (a), 6AGB/EMD-9616 and 6AH3/EMD-9622 (b), and 6EU2/EMD-3957 and 6EU1/EMD-3956 (c), respectively, plotted against the residue IDs, numbered according to their order in the coordinate files. r denotes the correlation coefficient between DynamicsMD and DynamicsDEFMap.

Extended Data Fig. 7 Atomic models of residues showing allosteric dynamics changes induced by ligand binding in the DMS3-RDM1 complex with DRD1 peptide.

a,b, The main chains of the hinge domain in DMS3 dimer 1 (a) and in the RDM1-DMS3 dimer 2 binding interface (b) are coloured according to the extent of ligand-induced changes in dynamics using the indicated colour bar. The side chains of residues showing the dynamics changes of less than -0.5 are represented as sticks (apo, magenta; holo, cyan).

Extended Data Fig. 8 Conformational changes of the spike protein of SARS-CoV-2.

a, Overall architecture of the spike protein of SARS-CoV-2. The spike protein mediates the entry of the virus into host cells. The protein is composed of two subunits: S1 subunit is involved in the recognition of the host cell receptor and S2 subunit is involved in the viral and membrane fusion process. The receptor binding domain in the S1 subunit exhibits a hinge-like motion and the ‘down’ and ‘up’ conformations are shown in grey and magenta, respectively. The conformational transition from the ‘down’ to ‘up’ states is indicated by a red arrow in the right-hand panel. The structural components associated with the structural transition from prefusion to post-fusion forms, the central helix, β-rich region, and the heptad-repeat 1 (HR1) motif, are indicated. b, Conformational rearrangement of the HR1 motif in the state transition from the prefusion to post-fusion forms. The post-fusion form of the spike protein in SARS-CoV-2 is shown in green by superimposing with the prefusion form. The conformational rearrangement of the HR1 motif is indicated by a yellow arrow. The prefusion form is coloured according to the indicated colour bar.

Extended Data Fig. 9 A correlation plot between DynamicsDEFMap calculated using EMD21375 and EMD-21375.

The regression lines are coloured orange; r denotes the correlation coefficient.

Extended Data Fig. 10 DEFMap-based dynamics visualization of virus particles.

The DEFMap-derived dynamics data are mapped onto the cryo-EM maps of the viral particles by applying the icosahedral symmetry. The colour range is defined by the minimum and maximum values in each inference. The scale bar represents 50 Å and is indicated by black lines. EMDB/PDB entries are indicated. The atomic model corresponding to EMD-9053 (right) is not available.

Supplementary information

Supplementary Information

Supplementary Figs. 1–3 and Tables 1–5.

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Matsumoto, S., Ishida, S., Araki, M. et al. Extraction of protein dynamics information from cryo-EM maps using deep learning. Nat Mach Intell 3, 153–160 (2021). https://doi.org/10.1038/s42256-020-00290-y

Download citation

  • Received:

  • Accepted:

  • Published:

  • Version of record:

  • Issue date:

  • DOI: https://doi.org/10.1038/s42256-020-00290-y

This article is cited by

Search

Advanced search

Quick links

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing