Unified rational protein engineering with sequence-based deep representation learning

Article metrics


Rational protein engineering requires a holistic understanding of protein function. Here, we apply deep learning to unlabeled amino-acid sequences to distill the fundamental features of a protein into a statistical representation that is semantically rich and structurally, evolutionarily and biophysically grounded. We show that the simplest models built on top of this unified representation (UniRep) are broadly applicable and generalize to unseen regions of sequence space. Our data-driven approach predicts the stability of natural and de novo designed proteins, and the quantitative function of molecularly diverse mutants, competitively with the state-of-the-art methods. UniRep further enables two orders of magnitude efficiency improvement in a protein engineering task. UniRep is a versatile summary of fundamental protein features that can be applied across protein engineering informatics.

Access options

Rent or Buy article

Get time limited or full article access on ReadCube.


All prices are NET prices.

Fig. 1: Workflow to learn and apply deep protein representations.
Fig. 2: UniRep encodes amino-acid physicochemistry, organism level information, secondary structure, evolutionary and functional information, and higher-order structural features.
Fig. 3: UniRep predicts structural and functional properties of proteins.
Fig. 4: UniRep, fine-tuned to a local evolutionary context, facilitates protein engineering by enabling generalization to distant peaks in the sequence landscape.

Data availability

All data are available in the main text or the supplementary materials.

Code availability

Code for UniRep model training and inference with trained weights along with links to all necessary data is available in a public repository at https://github.com/churchlab/UniRep. Code to reproduce all analysis and regenerate figures with links to preprocessed benchmark datasets is available online https://github.com/churchlab/UniRep-analysis.


  1. 1.

    Packer, M. S. & Liu, D. R. Methods for the directed evolution of proteins. Nat. Rev. Genet. 16, 379–394 (2015).

  2. 2.

    Romero, P. A. & Arnold, F. H. Exploring protein fitness landscapes by directed evolution. Nat. Rev. Mol. Cell Biol. 10, 866–876 (2009).

  3. 3.

    Biswas, S. et al. Toward machine-guided design of proteins. Preprint at bioRxiv https://doi.org/10.1101/337154 (2018).

  4. 4.

    Bedbrook, C. N., Yang, K. K., Rice, A. J., Gradinaru, V. & Arnold, F. H. Machine learning to design integral membrane channelrhodopsins for efficient eukaryotic expression and plasma membrane localization. PLoS Comput. Biol. 13, e1005786 (2017).

  5. 5.

    Rocklin, G. J. et al. Global analysis of protein folding using massively parallel design, synthesis, and testing. Science 357, 168–175 (2017).

  6. 6.

    Huang, P.-S., Boyken, S. E. & Baker, D. The coming of age of de novo protein design. Nature 537, 320–327 (2016).

  7. 7.

    Coluzza, I. Computational protein design: a review. J. Phys. Condens. Matter 29, 143001 (2017).

  8. 8.

    Romero, P. A., Krause, A. & Arnold, F. H. Navigating the protein fitness landscape with Gaussian processes. Proc. Natl Acad. Sci. USA 110, E193–E201 (2013).

  9. 9.

    Fox, R. J. et al. Improving catalytic function by ProSAR-driven enzyme evolution. Nat. Biotechnol. 25, 338 (2007).

  10. 10.

    Rohl, C. A., Strauss, C. E. M., Misura, K. M. S. & Baker, D. Protein structure prediction using rosetta. Numer. Computer Methods D. 383, 66–93 (2004).

  11. 11.

    Karplus, M. & Andrew McCammon, J. Molecular dynamics simulations of biomolecules. Nat. Struct. Mol. Biol. 9, 646 (2002).

  12. 12.

    Simon, J. R., Carroll, N. J., Rubinstein, M., Chilkoti, A. & López, G. P. Programming molecular self-assembly of intrinsically disordered proteins containing sequences of low complexity. Nat. Chem. 9, 509–515 (2017).

  13. 13.

    Taylor, N. D. et al. Engineering an allosteric transcription factor to respond to new ligands. Nat. Methods 13, 177–183 (2016).

  14. 14.

    Juárez, J. F., Lecube-Azpeitia, B., Brown, S. L., Johnston, C. D. & Church, G. M. Biosensor libraries harness large classes of binding domains for construction of allosteric transcriptional regulators. Nat. Commun. 9, 3101 (2018).

  15. 15.

    Riesselman, A. J., Ingraham, J. B. & Marks, D. S. Deep generative models of genetic variation capture the effects of mutations. Nat. Methods 15, 816–822 (2018).

  16. 16.

    AlQuraishi, M. End-to-end differentiable learning of protein structure. Cell Syst. 8, 292–301 (2019).

  17. 17.

    Liu, X. Deep recurrent neural network for protein function prediction from sequence. Preprint at arXiv https://arxiv.org/abs/1701.08318 (2017).

  18. 18.

    Schwartz, A. S. et al. Deep semantic protein representation for annotation, discovery, and engineering. Preprint at bioRxiv https://doi.org/10.1101/365965 (2018).

  19. 19.

    UniProtKB/TrEMBL 2018_10 (UniProt, accessed 21 November 2018); https://www.uniprot.org/statistics/TrEMBL

  20. 20.

    Asgari, E. & Mofrad, M. R. K. Continuous distributed representation of biological sequences for deep proteomics and genomics. PLoS ONE 10, e0141287 (2015).

  21. 21.

    Yang, K. K., Wu, Z., Bedbrook, C. N. & Arnold, F. H. Learned protein embeddings for machine learning. Bioinformatics 34, 2642–2648 (2018).

  22. 22.

    Suzek, B. E., Wang, Y., Huang, H., McGarvey, P. B. & Wu, C. H. UniRef clusters: a comprehensive and scalable alternative for improving sequence similarity searches. Bioinformatics 31, 926–932 (2015).

  23. 23.

    Radford, A., Jozefowicz, R. & Sutskever, I. Learning to generate reviews and discovering sentiment. Preprint at arXiv https://arxiv.org/abs/1704.01444 (2017).

  24. 24.

    van der Maaten, L. & Hinton, G. Visualizing data using t-SNE. J. Mach. Learn. Res. 37, 339–351 (2008).

  25. 25.

    Mizuguchi, K., Deane, C. M., Blundell, T. L. & Overington, J. P. HOMSTRAD: a database of protein structure alignments for homologous families. Protein Sci. 7, 2469–2471 (1998).

  26. 26.

    Raghava, G. P. S., Searle, S. M. J., Audley, P. C., Barber, J. D. & Barton, G. J. OXBench: a benchmark for evaluation of protein multiple sequence alignment accuracy. BMC Bioinforma. 4, 47 (2003).

  27. 27.

    Doan, A., Halevy, A. & Ives, Z. in Principles of Data Integration 95–119 (Elsevier, 2012).

  28. 28.

    Chua, S.-L. & Foo, L. K. Tree alignment based on Needleman–Wunsch algorithm for sensor selection in smart homes. Sensors 17, 1902 (2017).

  29. 29.

    Kwon, W. S., Da Silva, N. A. & Kellis, J. T. Jr. Relationship between thermal stability, degradation rate and expression yield of barnase variants in the periplasm of Escherichia coli. Protein Eng. 9, 1197–1202 (1996).

  30. 30.

    Bommarius, A. S. & Paye, M. F. Stabilizing biocatalysts. Chem. Soc. Rev. 42, 6534–6565 (2013).

  31. 31.

    Manning, M. C., Chou, D. K., Murphy, B. M., Payne, R. W. & Katayama, D. S. Stability of protein pharmaceuticals: an update. Pharm. Res. 27, 544–575 (2010).

  32. 32.

    Ovchinnikov, S. et al. Large-scale determination of previously unsolved protein structures using evolutionary information. eLife 4, e09248 (2015).

  33. 33.

    De novo designed protein AND identity:0.5 in UniRef (UnitProt, accessed 2 November 2018); https://www.uniprot.org/uniref/?query=de+novo+designed+protein+AND+identity%3A0.5

  34. 34.

    Quan, L., Lv, Q. & Zhang, Y. STRUM: structure-based prediction of protein stability changes upon single-point mutation. Bioinformatics 32, 2936–2946 (2016).

  35. 35.

    Gray, V. E., Hause, R. J., Luebeck, J., Shendure, J. & Fowler, D. M. Quantitative missense variant effect prediction using large-scale mutagenesis data. Cell Syst. 6, 116–124 (2018).

  36. 36.

    Zhang, C., Bengio, S., Hardt, M., Recht, B. & Vinyals, O. Understanding deep learning requires rethinking generalization. Preprint at arXiv https://arxiv.org/abs/1611.03530 (2016).

  37. 37.

    Sarkisyan, K. S. et al. Local fitness landscape of the green fluorescent protein. Nature 533, 397–401 (2016).

  38. 38.

    Rodriguez, E. A. et al. The growing and glowing toolbox of fluorescent and photoactive proteins. Trends Biochem. Sci. 42, 111–129 (2017).

  39. 39.

    Lambert, T. Tlambert03/Fpbase v.1.1.0 (Zenodo, 2018); https://doi.org/10.5281/ZENODO.1244328

  40. 40.

    Usmanova, D. R., Ferretti, L., Povolotskaya, I. S., Vlasov, P. K. & Kondrashov, F. A. A model of substitution trajectories in sequence space and long-term protein evolution. Mol. Biol. Evol. 32, 542–554 (2015).

  41. 41.

    Breen, M. S., Kemena, C., Vlasov, P. K., Notredame, C. & Kondrashov, F. A. Epistasis as the primary factor in molecular evolution. Nature 490, 535–538 (2012).

  42. 42.

    Dou, J. et al. De novo design of a fluorescence-activating β-barrel. Nature 561, 485–491 (2018).

  43. 43.

    Brookes, D. H., Park, H. & Listgarten, J. Conditioning by adaptive sampling for robust design. Proc. Machine Learn. Res. 97, 773–782 (2019).

  44. 44.

    Snoek, J. et al. Scalable Bayesian optimization using deep neural networks. Preprint at arXiv https://arxiv.org/abs/1502.05700 (2015).

  45. 45.

    Hernández-Lobato, J. M., Requeima, J., Pyzer-Knapp, E. O. & Aspuru-Guzik, A. Parallel and distributed thompson sampling for large-scale accelerated exploration of chemical space. Preprint at arXiv https://arxiv.org/abs/1706.01825 (2017).

  46. 46.

    Snoek, J., Larochelle, H. & Adams, R. P. in Advances in Neural Information Processing Systems Vol. 25 (eds. Pereira, F. et al.) 2951–2959 (Curran Associates, Inc., 2012).

  47. 47.

    Griffiths, R.-R. & Hernández-Lobato, J. M. Constrained Bayesian optimization for automaticchemical design. Preprint at arXiv https://arxiv.org/abs/1709.05501 (2017).

  48. 48.

    Gómez-Bombarelli, R. et al. Automatic chemical design using a data-driven continuous representation of molecules. ACS Cent. Sci. 4, 268–276 (2018).

  49. 49.

    Yang, K. K., Chen, Y., Lee, A. & Yue, Y. Batched stochastic Bayesian optimization via combinatorial constraints design. Preprint at arXiv https://arxiv.org/abs/1904.08102 (2019).

  50. 50.

    González, J., Longworth, J., James, D. C. & Lawrence, N. D. Bayesian optimization for synthetic gene design. Preprint at arXiv https://arxiv.org/abs/1505.01627 (2015).

  51. 51.

    Rinke, C. et al. Insights into the phylogeny and coding potential of microbial dark matter. Nature 499, 431–437 (2013).

  52. 52.

    Parks, D. H. et al. Recovery of nearly 8,000 metagenome-assembled genomes substantially expands the tree of life. Nat. Microbiol. 2, 1533 (2017).

  53. 53.

    EMBL-EBI. Current Release Statistics (UniProt, accessed 1 November 2018); https://www.ebi.ac.uk/uniprot/TrEMBLstats

  54. 54.

    Jouppi, N. P. et al. In-datacenter performance analysis of a tensorprocessing unit. In Proc. 44th Annual International Symposium of Computer Architecture Vol. 45, 1–12 (ACM, 2017).

  55. 55.

    Plesa, C., Sidore, A. M., Lubock, N. B., Zhang, D. & Kosuri, S. Multiplexed gene synthesis in emulsions for exploring protein functional landscapes. Science 359, 343–347 (2018).

  56. 56.

    Gu, L. et al. Multiplex single-molecule interaction profiling of DNA-barcoded proteins. Nature 515, 554–557 (2014).

  57. 57.

    Nutiu, R. et al. Direct measurement of DNA affinity landscapes on a high-throughput sequencing instrument. Nat. Biotechnol. 29, 659–664 (2011).

  58. 58.

    Thompson, D. B. et al. The future of multiplexed eukaryotic genome engineering. ACS Chem. Biol. 13, 313–325 (2018).

  59. 59.

    Ruder, S. An overview of multi-task learning in deep neural networks. Preprint at arXiv https://arxiv.org/abs/1706.05098 (2017).

  60. 60.

    Fox, N. K., Brenner, S. E. & Chandonia, J.-M. SCOPe: structural classification of proteins-extended, integrating SCOP and ASTRAL data and classification of new structures. Nucleic Acids Res 42, D304–D309 (2014).

  61. 61.

    Krause, B., Lu, L., Murray, I. & Renals, S. Multiplicative LSTM for sequence modelling. Preprint at arXiv https://arxiv.org/abs/1609.07959 (2016).

  62. 62.

    Gers, F. A., Schmidhuber, J. & Cummins, F. Learning to forget: continual prediction with LSTM. Neural Comput. 12, 2451–2471 (2000).

  63. 63.

    Cho, K., van Merrienboer, B., Bahdanau, D. & Bengio, Y. On the properties of neural machine translation: encoder-decoder approaches. In Eighth Workshop on Syntax, Semantics and Structure in Statistical Translation (2014).

  64. 64.

    Salimans, T. & Kingma, D. P. Weight normalization: a simple reparameterization to accelerate training of deep neural networks. Preprint at arXiv https://arxiv.org/abs/1602.07868 (2016).

  65. 65.

    AlQuraishi, M. ProteinNet: a standardized data set for machine learning of protein structure. BMC Bioinform. 20, 311 (2019).

  66. 66.

    Robertson, S. Understanding inverse document frequency: on theoretical arguments for IDF. J. Documentation 60, 503–520 (2004).

  67. 67.

    Park, H. et al. Simultaneous optimization of biomolecular energy functions on features from small molecules and macromolecules. J. Chem. Theory Comput. 12, 6201–6212 (2016).

  68. 68.

    Alford, R. F. et al. The rosetta all-atom energy function for macromolecular modeling and design. J. Chem. Theory Comput. 13, 3031–3048 (2017).

  69. 69.

    Glorot, X., Bordes, A. & Bengio, Y. Domain adaptation for large-scale sentiment classification: a deep learning approach. In Proc. 28th International Conference on International Conference on Machine Learning 513–520 (Omnipress, 2011).

  70. 70.

    Håndstad, T., Hestnes, A. J. H. & Sætrom, P. Motif kernel generated by genetic programming improves remote homology and fold detection. BMC Bioinform. 8, 23 (2007).

  71. 71.

    Li, S., Chen, J. & Liu, B. Protein remote homology detection based on bidirectional long short-term memory. BMC Bioinform. 18, 443 (2017).

  72. 72.

    Lovato, P., Cristani, M. & Bicego, M. Soft Ngram representation and modeling for protein remote homology detection. IEEE/ACM Trans. Comput. Biol. Bioinform. 14, 1482–1488 (2017).

  73. 73.

    Pedregosa, F. et al. Scikit-learn: machine learning in python. J. Mach. Learn. Res. 12, 2825–2830 (2011).

  74. 74.

    Jones, E., Oliphant, T. & Peterson, P. SciPy: Open source scientific tools for Python (SciPy, 2001); http://www.scipy.org/

  75. 75.

    2.3. Clustering—scikit-learn 0.20.0 documentation (scikit, 2018); http://scikit-learn.org/stable/modules/clustering.html

  76. 76.

    Alieva, N. O. et al. Diversity and evolution of coral fluorescent proteins. PLoS ONE 3, e2680 (2008).

  77. 77.

    EMBL-EBI, H. jackhmmer search | HMMER (EBI, accessed 2 November 2018); https://www.ebi.ac.uk/Tools/hmmer/search/jackhmmer

  78. 78.

    Thompson, J. D., Gibson, T. J. & Higgins, D. G. Multiple sequence alignment using ClustalW and ClustalX. Curr. Protoc. Bioinforma. 2, 2.3.1–2.3.22 (2002).

  79. 79.

    Zdobnov, E. M. et al. OrthoDBv9.1: cataloging evolutionary and functional annotations for animal, fungal, plant, archaeal, bacterial and viral orthologs. Nucleic Acids Res. 45, D744–D749 (2017).

  80. 80.

    Berman, H. M. et al. The Protein Data Bank. Nucleic Acids Res. 28, 235–242 (2000).

  81. 81.

    Kabsch, W. & Sander, C. Dictionary of protein secondary structure: pattern recognition of hydrogen-bonded and geometrical features. Biopolym.: Original Res. Biomolecules 22, 2577–2637 (1983).

  82. 82.

    Alley E. et al. Unified rational protein engineering with sequence-based deep representation learning protocol. Preprint at bioRxiv https://doi.org/10.1101/589333 (2019).

Download references


We thank J. Aach, A. Taylor-Weiner, D. Goodman, P. Ogden, G. Kuznetsov, S. Sinai, A. Tucker, M. Turpin, J. Swett, N. Thomas, R. Sha, C. Bakerlee and K. Fish for valuable feedback and discussion. S.B. was supported by an NIH Training Grant (no. T32HG002295) to the Harvard Bioinformatics and Integrative Genomics program as well as an NSF GRFP Fellowship. M.A. was supported through NIGMS Grant no. P50GM107618 and NIH grant no. U54-CA225088. E.C.A. and G.K. were supported by the Center for Effective Altruism. E.C.A. was partially supported by the Wyss Institute for Biologically Inspired Engineering. Computational resources were, in part, generously provided by the AWS Cloud Credits for the Research program.

Author information

E.C.A. and G.K. conceived the study. E.C.A., G.K. and S.B. conceived the experiments, managed data and performed the analysis. M.A. performed large-scale RGN inference and managed data and software for parts of the analysis. G.M.C. supervised the project. E.C.A., G.K. and S.B. wrote the manuscript with help from all authors.

Correspondence to George M. Church.

Ethics declarations

Competing interests

E.C.A., G.K. and S.B. are in the process of pursuing a patent on this technology. S.B. is a former consultant for Flagship Pioneering company VL57 (now VL56). A full list of G.M.C.’s tech transfer, advisory roles and funding sources can be found on the laboratory’s website: http://arep.med.harvard.edu/gmc/tech.html.

Additional information

Peer review information Nicole Rusk was the primary editor on this article and managed its editorial process and peer review in collaboration with the rest of the editorial team.

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

Supplementary information

Supplementary Information

Supplementary Figs. 1–15 and Tables 1–9.

Reporting Summary

Supplementary Data Set 1

All test set results in .xlsx format

Supplementary Data Set 2

All validation set results in .xlsx format

Supplementary Data Set 3

All test set results graphically presented in .pdf format

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Alley, E.C., Khimulya, G., Biswas, S. et al. Unified rational protein engineering with sequence-based deep representation learning. Nat Methods (2019) doi:10.1038/s41592-019-0598-1

Download citation