Deep learning extends de novo protein modelling coverage of genomes using iteratively predicted structural constraints

Abstract

The inapplicability of amino acid covariation methods to small protein families has limited their use for structural annotation of whole genomes. Recently, deep learning has shown promise in allowing accurate residue-residue contact prediction even for shallow sequence alignments. Here we introduce DMPfold, which uses deep learning to predict inter-atomic distance bounds, the main chain hydrogen bond network, and torsion angles, which it uses to build models in an iterative fashion. DMPfold produces more accurate models than two popular methods for a test set of CASP12 domains, and works just as well for transmembrane proteins. Applied to all Pfam domains without known structures, confident models for 25% of these so-called dark families were produced in under a week on a small 200 core cluster. DMPfold provides models for 16% of human proteome UniProt entries without structures, generates accurate models with fewer than 100 sequences in some cases, and is freely available.

Full text

Its licence does not allow us to show the full text. Read it at PubMed Central:

References 49

1. 1.

   de Juan D, Pazos F, Valencia A. Emerging methods in protein co-evolution. Nat. Rev. Genet. 2013;14:249-261.

   Not in our library10.1038/nrg3414 (opens in a new tab)

2. 2.

   Monastyrskyy B, D’Andrea D, Fidelis K, Tramontano A, Kryshtafovych A. New encouraging developments in contact prediction: Assessment of the CASP11 results. Protein. Struct. Funct. Bioinf. 2015;84:131-144.

   Not in our library10.1002/prot.24943 (opens in a new tab)

3. 3.

   Michel M, et al. PconsFold: improved contact predictions improve protein models. Bioinformatics. 2014;30:i482-i488.

   Not in our library10.1093/bioinformatics/btu458 (opens in a new tab)

4. 4.

   Bender BJ, et al. Protocols for molecular modeling with Rosetta3 and RosettaScripts. Biochemistry. 2016;55:4748-4763.

   Not in our library10.1021/acs.biochem.6b00444 (opens in a new tab)

5. 5.

   Kosciolek T, Jones DT. De novo structure prediction of globular proteins aided by sequence variation-derived contacts. PLoS ONE. 2014;9:e92197.

   Not in our library10.1371/journal.pone.0092197 (opens in a new tab)

6. 6.

   Ovchinnikov S, Park H, Kim DE, DiMaio F, Baker D. Protein structure prediction using Rosetta in CASP12. Protein. Struct. Funct. Bioinf. 2018;86:113-121.

   Not in our library10.1002/prot.25390 (opens in a new tab)

7. 7.

   Kim DE, Chivian D, Malmström L, Baker D. Automated prediction of domain boundaries in CASP6 targets using Ginzu and RosettaDOM. Protein. Struct. Funct. Bioinf. 2005;61:193-200.

   Not in our library10.1002/prot.20737 (opens in a new tab)

8. 8.

   Aszódi A, Gradwell MJ, Taylor WR. Global fold determination from a small number of distance restraints. J. Mol. Biol. 1995;251:308-326.

   Not in our library10.1006/jmbi.1995.0436 (opens in a new tab)

9. 9.
   CONFOLD2: improved contact-driven ab initio protein structure modeling · 2018

   Not checked
10. 10.
    Protein 3D Structure Computed from Evolutionary Sequence Variation · 2011

    No notice
11. 11.

    Brunger AT. Version 1.2 of the crystallography and NMR system. Nat. Protoc. 2007;2:2728-2733.

    Not in our library10.1038/nprot.2007.406 (opens in a new tab)

12. 12.

    Pastore A, Atkinson RA, Saudek V, Williams RJ. Topological mirror images in protein structure computation: an underestimated problem. Protein. Struct. Funct. Bioinf. 1991;10:22-32.

    Not in our library10.1002/prot.340100104 (opens in a new tab)

13. 13.

    Michel M, Menéndez Hurtado D, Uziela K, Elofsson A. Large-scale structure prediction by improved contact predictions and model quality assessment. Bioinformatics. 2017;33:i23-i29.

    Not in our library10.1093/bioinformatics/btx239 (opens in a new tab)

14. 14.

    Xu Jinbo. Distance-based protein folding powered by deep learning. Proceedings of the National Academy of Sciences. 2019;116(34):16856-16865.

    Not in our library10.1073/pnas.1821309116 (opens in a new tab)

15. 15.

    AlQuraishi M. End-to-end differentiable learning of protein structure. Cell Syst. 2019;8:292-301.

    Not in our library10.1016/j.cels.2019.03.006 (opens in a new tab)

16. 16.

    Anonymous. Learning protein structure with a differentiable simulator. ICLR 2019 Conference Blind Submission. (2018).

    Not in our library

17. 17.

    Kandathil S. M., Greener J. G., Jones D. T. Prediction of interresidue contacts with DeepMetaPSICOV in CASP13. Proteins Struct. Funct. Bioinf. (2019).

    Not in our library10.1002/prot.25779 (opens in a new tab)

18. 18.
    The Pfam protein families database in 2019 · 2018

    No notice
19. 19.

    Ovchinnikov S, et al. Protein structure determination using metagenome sequence data. Science. 2017;355:294-298.

    Not in our library10.1126/science.aah4043 (opens in a new tab)

20. 20.

    Lamb John, Jarmolinska Aleksandra I., Michel Mirco, Menéndez-Hurtado David, Sulkowska Joanna I., Elofsson Arne. PconsFam: An Interactive Database of Structure Predictions of Pfam Families. Journal of Molecular Biology. 2019;431(13):2442-2448.

    Not in our library10.1016/j.jmb.2019.01.047 (opens in a new tab)

21. 21.

    Schaeffer RD, Liao Y, Cheng H, Grishin NV. ECOD: new developments in the evolutionary classification of domains. Nucleic Acids Res. 2017;45:D296-D302.

    Not in our library10.1093/nar/gkw1137 (opens in a new tab)

22. 22.

    Zhang Y, Skolnick J. Scoring function for automated assessment of protein structure template quality. Protein. Struct. Funct. Bioinf. 2004;57:702-710.

    Not in our library10.1002/prot.20264 (opens in a new tab)

23. 23.

    Kabsch, W. & Sander, C. Dictionary of protein secondary structure: Pattern recognition of hydrogen-bonded and geometrical features. Biopolymers22, 2577–2637 (1983).

    Not in our library10.1002/bip.360221211 (opens in a new tab)

24. 24.

    Hopf TA, et al. Three-dimensional structures of membrane proteins from genomic sequencing. Cell. 2012;149:1607-1621.

    Not in our library10.1016/j.cell.2012.04.012 (opens in a new tab)

25. 25.

    Nugent T, Jones DT. Accurate de novo structure prediction of large transmembrane protein domains using fragment-assembly and correlated mutation analysis. Proc. Natl Acad. Sci. USA. 2012;109:E1540-E1547.

    Not in our library10.1073/pnas.1120036109 (opens in a new tab)

26. 26.

    Perdigão N, et al. Unexpected features of the dark proteome. Proc. Natl Acad. Sci. USA. 2015;112:15898-15903.

    Not in our library10.1073/pnas.1508380112 (opens in a new tab)

27. 27.

    Dessailly BH, et al. PSI-2: structural genomics to cover protein domain family space. Structure. 2009;17:869-881.

    Not in our library10.1016/j.str.2009.03.015 (opens in a new tab)

28. 28.

    Jones DT, Buchan DWA, Cozzetto D, Pontil M. PSICOV: precise structural contact prediction using sparse inverse covariance estimation on large multiple sequence alignments. Bioinformatics. 2012;28:184-190.

    Not in our library10.1093/bioinformatics/btr638 (opens in a new tab)

29. 29.

    Seemayer S, Gruber M, Söding J. CCMpred–fast and precise prediction of protein residue-residue contacts from correlated mutations. Bioinformatics. 2014;30:3128-3130.

    Not in our library10.1093/bioinformatics/btu500 (opens in a new tab)

30. 30.

    Lewis TE, et al. Genome3D: exploiting structure to help users understand their sequences. Nucleic Acids Res. 2015;43:D382-D386.

    Not in our library10.1093/nar/gku973 (opens in a new tab)

31. 31.
    UniProt: the universal protein knowledgebase · 2018

    No notice
32. 32.

    Somody JC, MacKinnon SS, Windemuth A. Structural coverage of the proteome for pharmaceutical applications. Drug Disco. Today. 2017;22:1792-1799.

    Not in our library10.1016/j.drudis.2017.08.004 (opens in a new tab)

33. 33.

    Jumper, J. De novo protein folding using statistical potentials from deep learning. CASP13 presentation (2018).

    Not in our library

34. 34.

    Jones DT, Kandathil SM. High precision in protein contact prediction using fully convolutional neural networks and minimal sequence features. Bioinformatics. 2018;34:3308-3315.

    Not in our library10.1093/bioinformatics/bty341 (opens in a new tab)

35. 35.

    Goodfellow I, Warde-Farley D, Mirza M, Courville A, Bengio Y. Maxout networks. PMLR. 2013;28:1319-1327.

    Not in our library

36. 36.

    Kingma, D. P., Ba, J. Adam: a method for stochastic optimization. Proceedings of the 3rd International Conference on Learning Representations (ICLR) (2015).

    Not in our library

37. 37.

    Remmert M, Biegert A, Hauser A, Söding J. HHblits: lightning-fast iterative protein sequence searching by HMM-HMM alignment. Nat. Methods. 2011;9:173-175.

    Not in our library10.1038/nmeth.1818 (opens in a new tab)

38. 38.

    Hochreiter S, Schmidhuber J. Long short-term memory. Neural Comput. 1997;9:1735-1780.

    Not in our library10.1162/neco.1997.9.8.1735 (opens in a new tab)

39. 39.

    Heffernan R, Yang Y, Paliwal K, Zhou Y. Capturing non-local interactions by long short-term memory bidirectional recurrent neural networks for improving prediction of protein secondary structure, backbone angles, contact numbers and solvent accessibility. Bioinformatics. 2017;33:2842-2849.

    Not in our library10.1093/bioinformatics/btx218 (opens in a new tab)

40. 40.

    Li, H., Hou, J., Adhikari, B., Lyu, Q., Cheng, J. Deep learning methods for protein torsion angle prediction. BMC Bioinform. 18, 417 (2017).

    Not in our library10.1186/s12859-017-1834-2 (opens in a new tab)

41. 41.

    Pettitt CS, McGuffin LJ, Jones DT. Improving sequence-based fold recognition by using 3D model quality assessment. Bioinformatics. 2005;21:3509-3515.

    Not in our library10.1093/bioinformatics/bti540 (opens in a new tab)

42. 42.

    Webb B, Sali A. Comparative protein structure modeling using MODELLER. Curr. Protoc. Protein Sci. 2016;54:5.6.1-5.6.37.

    Not in our library10.1002/cpbi.3 (opens in a new tab)

43. 43.
    CD-HIT: accelerated for clustering the next-generation sequencing data · 2012

    No notice
44. 44.

    Jones DT. Protein secondary structure prediction based on position-specific scoring matrices. J. Mol. Biol. 1999;292:195-202.

    Not in our library10.1006/jmbi.1999.3091 (opens in a new tab)

45. 45.

    Zimmermann L, et al. A completely reimplemented MPI bioinformatics toolkit with a new HHpred server at its core. J. Mol. Biol. 2018;430:2237-2243.

    Not in our library10.1016/j.jmb.2017.12.007 (opens in a new tab)

46. 46.

    Potter SC, et al. HMMER web server: 2018 update. Nucleic Acids Res. 2018;46:W200-W204.

    Not in our library10.1093/nar/gky448 (opens in a new tab)

47. 47.

    Zhang Y, Skolnick J. TM-align: a protein structure alignment algorithm based on the TM-score. Nucleic Acids Res. 2005;33:2302-2309.

    Not in our library10.1093/nar/gki524 (opens in a new tab)

48. 48.

    Klambauer, G., Unterthiner, T., May, A. Self-normalizing neural networks. Adv. Neural. Inf. Process. Syst.30, 971–980 (2017).

    Not in our library

49. 49.

    Paszke, A. et al. Automatic differentiation in PyTorch. NIPS Autodiff Workshophttps://github.com/pytorch/pytorch/blob/master/CITATION (2017).

    Not in our library

Cited by papers we hold 7

• Modeling protein conformational ensembles by guiding AlphaFold2 with Double Electron Electron Resonance (DEER) distance distributions · 2025
  Not checked
• Protein remote homology detection and structural alignment using deep learning · 2023
  No notice
• Improved Protein Structure Prediction Using a New Multi‐Scale Network and Homologous Templates · 2021
  Not checked
• Structure-based protein function prediction using graph convolutional networks · 2021
  No notice
• Highly accurate protein structure prediction for the human proteome · 2021
  No notice
• PolyFold: An interactive visual simulator for distance-based protein folding. · 2020
  Not checked
• Predicting changes in protein thermodynamic stability upon point mutation with deep 3D convolutional neural networks · 2020
  No notice

Questions about this paper

Reads

A person counts once a day per paper. We keep no cookie and no address: a reader is a hash that changes every day. Any month under five reads shows no number.