Nature research paper: Rational design of disordered proteins for sequence–function investigation

https://t.co/BRWg6DS0Er

Rational design of disordered proteins for sequence–function investigation

Source: https://www.nature.com/articles/s41586-026-10849-1?utm_source=x&utm_medium=social&utm_campaign=nature&linkId=63004835&error=cookies_not_supported&code=6960ef02-92f3-4072-a214-e565079b6853

Data availability

Data and analysis scripts used for figures and analysis in this paper are available at GitHub (https://github.com/sukeniklab/GOOSE_2026andhttps://github.com/holehouse-lab/supportingdata/tree/master/2026/GOOSE_2026). Raw sequencing data have been deposited at Zenodo82(https://zenodo.org/records/18774462).

Code availability

GOOSE is fully open source and is available at GitHub (https://github.com/idptools/goose/). Documentation for using GOOSE is available online (https://goose.readthedocs.io/en/latest/). A subset of GOOSE’s functionality is available across two Colab Notebooks. For sequence generation and variant design, use the following notebook:https://colab.research.google.com/drive/1U9B-TfoNEZbbjhPUG5lrMPS0JL0nDB3o?usp=sharing. For creating sequences with specific homotypic and/or heterotypic IDR–IDR interactions, use the following notebook:https://colab.research.google.com/drive/1aJajo1IK66ApFSMwCCBumCZixlfshw7A?usp=sharing.

References

1. Holehouse, A. S. & Kragelund, B. B. The molecular basis for cellular function of intrinsically disordered protein regions.*Nat. Rev. Mol. Cell Biol.*25, 187–211 (2024). ArticlePubMedGoogle Scholar
2. Albanese, K. I., Barbe, S., Tagami, S., Woolfson, D. N. & Schiex, T. Computational protein design.Nat. Rev. Methods Primers5, 13 (2025). ArticleGoogle Scholar
3. Tesei, G., Pesce, F. & Lindorff-Larsen, K. Computational design of intrinsically disordered proteins.*Curr. Opin. Struct. Biol.*96, 103210 (2026). ArticlePubMedGoogle Scholar
4. Moses, D. et al. Structural biases in disordered proteins are prevalent in the cell.*Nat. Struct. Mol. Biol.*31, 283–292 (2024). ArticlePubMedPubMed CentralGoogle Scholar
5. Das, R. K., Ruff, K. M. & Pappu, R. V. Relating sequence encoded information to form and function of intrinsically disordered proteins.*Curr. Opin. Struct. Biol.*32, 102–112 (2015). ArticlePubMedPubMed CentralGoogle Scholar
6. Das, R. K. & Pappu, R. V. Conformations of intrinsically disordered proteins are influenced by linear sequence distributions of oppositely charged residues.Proc. Natl Acad. Sci. USA110, 13392–13397 (2013). ArticleADSPubMedPubMed CentralGoogle Scholar
7. Staller, M. V. et al. A high-throughput mutational scan of an intrinsically disordered acidic transcriptional activation domain.*Cell Syst.*6, 444–455 (2018). ArticlePubMedPubMed CentralGoogle Scholar
8. Pesce, F. et al. Design of intrinsically disordered protein variants with diverse structural properties.*Sci. Adv.*10, eadm9926 (2024). ArticlePubMedPubMed CentralGoogle Scholar
9. Zarin, T., Tsai, C. N., Nguyen Ba, A. N. & Moses, A. M. Selection maintains signaling function of a highly diverged intrinsically disordered region.Proc. Natl Acad. Sci. USA114, E1450–E1459 (2017). ArticleADSPubMedPubMed CentralGoogle Scholar
10. Das, R. K., Huang, Y., Phillips, A. H., Kriwacki, R. W. & Pappu, R. V. Cryptic sequence features within the disordered protein p27Kip1 regulate cell cycle signaling.Proc. Natl Acad. Sci. USA113, 5616–5621 (2016). ArticleADSPubMedPubMed CentralGoogle Scholar
11. Holehouse, A. S., Das, R. K., Ahad, J. N., Richardson, M. O. G. & Pappu, R. V. CIDER: resources to analyze sequence-ensemble relationships of intrinsically disordered proteins.*Biophys. J.*112, 16–21 (2017). ArticleADSPubMedPubMed CentralGoogle Scholar
12. Novak, B., Lotthammer, J. M., Emenecker, R. J. & Holehouse, A. S. Accurate predictions of disordered protein ensembles with STARLING.Nature652, 240–250 (2026). ArticlePubMedPubMed CentralGoogle Scholar
13. González-Foutel, N. S. et al. Conformational buffering underlies functional selection in intrinsically disordered protein regions.*Nat. Struct. Mol. Biol.*29, 781–790 (2022). ArticlePubMedPubMed CentralGoogle Scholar
14. Zeno, W. F. et al. Molecular mechanisms of membrane curvature sensing by a disordered protein.*J. Am. Chem. Soc.*141, 10361–10371 (2019). ArticleADSPubMedPubMed CentralGoogle Scholar
15. Halladin, D. K. et al. Entropy-driven translocation of disordered proteins through the Gram-positive bacterial cell wall.*Nat. Microbiol.*6, 1055–1065 (2021). ArticlePubMedPubMed CentralGoogle Scholar
16. Berlow, R. B., Dyson, H. J. & Wright, P. E. Hypersensitive termination of the hypoxic response by a disordered protein switch.Nature543, 447–451 (2017). ArticleADSPubMedPubMed CentralGoogle Scholar
17. Schuler, B., König, I., Soranno, A. & Nettels, D. Impact of in-cell and in-vitro crowding on the conformations and dynamics of an intrinsically disordered protein.Angew. Chem. Int. Ed.https://doi.org/10.1002/anie.202016804(2021).
18. Ruff, K. M. et al Molecular grammars of predicted intrinsically disordered regions that span the human proteome.Cell189, 323–342 (2025).
19. Moses, D. et al. Revealing the hidden sensitivity of intrinsically disordered proteins to their chemical environment.*J. Phys. Chem. Lett.*11, 10131–10136 (2020). ArticlePubMedPubMed CentralGoogle Scholar
20. Marsh, J. A. & Forman-Kay, J. D. Sequence determinants of compaction in intrinsically disordered proteins.*Biophys. J.*98, 2383–2390 (2010). ArticleADSPubMedPubMed CentralGoogle Scholar
21. Müller-Späth, S. et al. From the cover: charge interactions can dominate the dimensions of intrinsically disordered proteins.Proc. Natl Acad. Sci. USA107, 14609–14614 (2010). ArticleADSPubMedPubMed CentralGoogle Scholar
22. Mao, A. H., Crick, S. L., Vitalis, A., Chicoine, C. L. & Pappu, R. V. Net charge per residue modulates conformational ensembles of intrinsically disordered proteins.Proc. Natl Acad. Sci. USA107, 8183–8188 (2010). ArticleADSPubMedPubMed CentralGoogle Scholar
23. Sørensen, C. S. & Kjaergaard, M. Effective concentrations enforced by intrinsically disordered linkers are governed by polymer physics.Proc. Natl Acad. Sci. USA116, 23124–23131 (2019). ArticleADSPubMedPubMed CentralGoogle Scholar
24. Boeynaems, S. et al. Aberrant phase separation is a common killing strategy of positively charged peptides in biology and human disease. Preprint atbioRxivhttps://doi.org/10.1101/2023.03.09.531820(2023).
25. Schmidt, H. B. & Görlich, D. Transport selectivity of nuclear pores, phase separation, and membraneless organelles.*Trends Biochem. Sci.*41, 46–61 (2016). ArticlePubMedGoogle Scholar
26. Enriquez-Toledo, C. et al. A molecular grammar for environmental sensitivity in intrinsically disordered protein regions. Preprint atbioRxivhttps://doi.org/10.64898/2026.04.20.719774(2026).
27. Soranno, A. et al. Single-molecule spectroscopy reveals polymer effects of disordered proteins in crowded environments.Proc. Natl Acad. Sci. USA111, 4874–4879 (2014). ArticleADSPubMedPubMed CentralGoogle Scholar
28. Sukenik, S., Ren, P. & Gruebele, M. Weak protein-protein interactions in live cells are quantified by cell-volume modulation.Proc. Natl Acad. Sci. USA114, 6776–6781 (2017). ArticleADSPubMedPubMed CentralGoogle Scholar
29. Speer, S. L., Stewart, C. J., Sapir, L., Harries, D. & Pielak, G. J. Macromolecular crowding is more than hard-core repulsions.Annu. Rev. Biophys.https://doi.org/10.1146/annurev-biophys-091321-071829(2022).
30. Ginell, G. M. et al. Sequence-based prediction of intermolecular interactions driven by disordered regions.Science388, eadq8381 (2025). ArticlePubMedPubMed CentralGoogle Scholar
31. Martin, E. W. et al. Valence and patterning of aromatic residues determine the phase behavior of prion-like domains.Science367, 694–699 (2020). ArticleADSPubMedPubMed CentralGoogle Scholar
32. Wang, J. et al. A molecular grammar governing the driving forces for phase separation of prion-like RNA binding proteins.Cell174, 688–699 (2018). ArticleADSPubMedPubMed CentralGoogle Scholar
33. Marks, R. A. et al. Life on the dry side: a roadmap to understanding desiccation tolerance and accelerating translational applications.*Nat. Commun.*16, 3284 (2025). ArticleADSPubMedPubMed CentralGoogle Scholar
34. Olgenblum, G. I., Hutcheson, B. O., Pielak, G. J. & Harries, D. Protecting proteins from desiccation stress using molecular glasses and gels.*Chem. Rev.*124, 5668–5694 (2024). ArticlePubMedPubMed CentralGoogle Scholar
35. Hibshman, J. D. & Goldstein, B. LEA motifs promote desiccation tolerance in vivo.*BMC Biol.*19, 263 (2021). ArticlePubMedPubMed CentralGoogle Scholar
36. Nguyen, K. et al. A phase transition modulates the protective function of a tardigrade disordered protein during desiccation.*Protein Sci.*34, e70300 (2025). ArticlePubMedPubMed CentralGoogle Scholar
37. Lotthammer, J. M., Ginell, G. M., Griffith, D., Emenecker, R. J. & Holehouse, A. S. Direct prediction of intrinsically disordered protein conformational properties from sequence.Nat. Methods21, 465–476 (2024). ArticlePubMedPubMed CentralGoogle Scholar
38. Tolleter, D. et al. Structure and function of a mitochondrial late embryogenesis abundant protein are revealed by desiccation.Plant Cell19, 1580–1589 (2007). ArticlePubMedPubMed CentralGoogle Scholar
39. Bremer, A., Wolff, M., Thalhammer, A. & Hincha, D. K. Folding of intrinsically disordered plant LEA proteins is driven by glycerol-induced crowding and the presence of membranes.*FEBS J.*284, 919–936 (2017). ArticlePubMedGoogle Scholar
40. Shraddha, K. C. et al. Disordered proteins interact with the chemical environment to tune their protective function during drying.eLife13, RP97231 (2024). ArticleGoogle Scholar
41. Nicholson, V., Maharjan, S., Sukenik, S. & Boothby, T. C. OsmoFold: a high-throughput tool for predicting the impact of osmolytes on protein structure.Biophys. J.https://doi.org/10.1016/j.bpj.2025.08.034(2025).
42. Malki, A. et al. Fibril structure of desiccation-protective tardigrade protein CAHS-8.*Angew. Chem. Int. Ed.*65, e19912 (2026). ArticleGoogle Scholar
43. Sanchez-Martinez, S. et al. Labile assembly of a tardigrade protein induces biostasis.*Protein Sci.*33, e4941 (2024). ArticlePubMedPubMed CentralGoogle Scholar
44. Romero-Perez, P. S., Dorone, Y., Flores, E., Sukenik, S. & Boeynaems, S. When phased without water: biophysics of cellular desiccation, from biomolecules to condensates.*Chem. Rev.*123, 9010–9035 (2023). ArticlePubMedPubMed CentralGoogle Scholar
45. Langstein-Skora, I. et al. Sequence and chemical specificity define the functional landscape of intrinsically disordered regions.*Nat. Cell Biol.*28, 323–337 (2026). ArticlePubMedPubMed CentralGoogle Scholar
46. Greig, J. A. et al. Arginine-enriched mixed-charge domains provide cohesion for nuclear speckle condensation.Mol. Cell77, 1237–1250 (2020). ArticleADSPubMedPubMed CentralGoogle Scholar
47. Romero-Pérez, P. S. et al. Protein surface chemistry encodes an adaptive tolerance to desiccation.*Cell Syst.*16, 101407 (2025). ArticlePubMedGoogle Scholar
48. Solomon, D. A. et al. C9ORF72-derived polyGR polypeptides disrupt passive nucleocytoplasmic transport by tuning protein affinity for the nuclear pore barrier. Preprint atbioRxivhttps://doi.org/10.64898/2026.03.16.711670(2026).
49. Alshareedah, I. et al. Sequence-specific interactions determine viscoelasticity and aging dynamics of protein condensates.*Nat. Phys.*20, 1482–1491 (2024). ArticlePubMedPubMed CentralGoogle Scholar
50. Battaglia, M., Olvera-Carrillo, Y., Garciarrubio, A., Campos, F. & Covarrubias, A. A. The enigmatic LEA proteins and other hydrophilins.*Plant Physiol.*148, 6–24 (2008). ArticlePubMedPubMed CentralGoogle Scholar
51. Tesei, G. et al. Conformational ensembles of the human intrinsically disordered proteome.Nature626, 897–904 (2024). ArticleADSPubMedGoogle Scholar
52. Pritišanac, I. et al. A functional map of the human intrinsically disordered proteome.Proc. Natl Acad. Sci. USA123, e2604562123 (2026).
53. Alamdari, S. et al. Protein generation with evolutionary diffusion: sequence is all you need. Preprint atbioRxivhttps://doi.org/10.1101/2023.09.11.556673(2023).
54. Strome, B., Elemam, K., Pritisanac, I., Forman-Kay, J. D. & Moses, A. M. Computational design of intrinsically disordered protein regions by matching bulk molecular properties. Preprint atbioRxivhttps://doi.org/10.1101/2023.04.28.538739(2023).
55. Krueger, R. K., Brenner, M. P. & Shrinivas, K. Generalized design of sequence–ensemble–function relationships for intrinsically disordered proteins.*Nat. Comput. Sci.*6, 512–523 (2025).
56. Bhat, S. et al. De novo design of peptide binders to conformationally diverse targets with contrastive language modeling.*Sci. Adv.*11, eadr8638 (2025). ArticlePubMedPubMed CentralGoogle Scholar
57. Chen, L. T. et al. Target sequence-conditioned design of peptide binders using masked language modeling.*Nat. Biotechnol.*44, 1002–1010 (2025).
58. An, Y., Webb, M. A. & Jacobs, W. M. Active learning of the thermodynamics-dynamics trade-off in protein condensates.*Sci. Adv.*10, eadj2448 (2024). ArticlePubMedPubMed CentralGoogle Scholar
59. Brown, C. J., Johnson, A. K., Dunker, A. K. & Daughdrill, G. W. Evolution and disorder.*Curr. Opin. Struct. Biol.*21, 441–446 (2011). ArticlePubMedPubMed CentralGoogle Scholar
60. Lin, Z. et al. Evolutionary-scale prediction of atomic-level protein structure with a language model.Science379, 1123–1130 (2023). ArticleADSMathSciNetPubMedGoogle Scholar
61. Liu, J. X. et al. Generative design of intrinsically disordered protein regions with IDiom. Preprint atbioRxivhttps://doi.org/10.64898/2026.04.10.717777(2026).
62. Joseph, J. A. et al. Physics-driven coarse-grained model for biomolecular phase separation with near-quantitative accuracy.*Nat. Comput. Sci.*1, 732–743 (2021). ArticlePubMedPubMed CentralGoogle Scholar
63. Tesei, G. & Lindorff-Larsen, K. Improved predictions of phase behaviour of intrinsically disordered proteins by tuning the interaction range.*Open Res. Eur.*2, 94 (2022). ArticlePubMedGoogle Scholar
64. Lotthammer, J. M. et al Metapredict enables accurate disorder prediction across the Tree of Life. Preprint atbioRxivhttps://doi.org/10.1101/2024.11.05.622168(2024).
65. Stringer, C., Wang, T., Michaelos, M. & Pachitariu, M. Cellpose: a generalist algorithm for cellular segmentation.Nat. Methods18, 100–106 (2021). ArticlePubMedGoogle Scholar
66. Thevenaz, P., Ruttimann, U. E. & Unser, M. A pyramid approach to subpixel registration based on intensity.*IEEE Trans. Image Process.*7, 27–41 (1998). ArticleADSPubMedGoogle Scholar
67. Abraham, M. J. et al. GROMACS: high performance molecular simulations through multi-level parallelism from laptops to supercomputers.SoftwareX1–2, 19–25 (2015). ArticleADSGoogle Scholar
68. Wang, L., Brasnett, C., Borges-Araújo, L., Souza, P. C. T. & Marrink, S. J. Martini3-IDP: improved Martini 3 force field for disordered proteins.*Nat. Commun.*16, 2874 (2025). ArticleADSPubMedPubMed CentralGoogle Scholar
69. Grünewald, F. et al. Polyply; a Python suite for facilitating simulations of macromolecules and nanomaterials.*Nat. Commun.*13, 68 (2022). ArticleADSPubMedPubMed CentralGoogle Scholar
70. Lambert, T. J. FPbase: a community-editable fluorescent protein database.Nat. Methods16, 277–278 (2019). ArticlePubMedGoogle Scholar
71. Joska, T. M., Mashruwala, A., Boyd, J. M. & Belden, W. J. A universal cloning method based on yeast homologous recombination that is simple, efficient, and versatile.J. Microbiol. Methods100, 46–51 (2014). ArticlePubMedPubMed CentralGoogle Scholar
72. Hood, S. & Lewis, J. A. High-efficiency yeast electroporation v.2.Protocols.iohttps://doi.org/10.17504/protocols.io.5qpvorr69v4o/v2(2022).
73. Calahan, D., Dunham, M., DeSevo, C. & Koshland, D. E. Genetic analysis of desiccation tolerance inSaccharomyces cerevisiae.Genetics189, 507–519 (2011). ArticlePubMedPubMed CentralGoogle Scholar
74. Stokes, R. H. & Robinson, R. A. Standard solutions for humidity control at 25 °C.*Ind. Eng. Chem.*41, 2013–2013 (1949). ArticleGoogle Scholar
75. Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2.*Genome Biol.*15, 550 (2014). ArticlePubMedPubMed CentralGoogle Scholar
76. Langmead, B. & Salzberg, S. L. Fast gapped-read alignment with Bowtie 2.Nat. Methods9, 357–359 (2012). ArticlePubMedPubMed CentralGoogle Scholar
77. Danecek, P. et al. Twelve years of SAMtools and BCFtools.Gigascience10, giab008 (2021). ArticlePubMedPubMed CentralGoogle Scholar
78. Li, H. et al. The Sequence Alignment/Map format and SAMtools.Bioinformatics25, 2078–2079 (2009). ArticlePubMedPubMed CentralGoogle Scholar
79. Dunham, M., Gartenberg, M. & Brown, G. W.Methods in Yeast Genetics and Genomics: A Cold Spring Harbor Laboratory Course Manual, 2015 Edition(Cold Spring Harbor Laboratory Press, 2015)
80. Li, C. H. & Lee, C. K. Minimum cross entropy thresholding.*Pattern Recognit.*26, 617–625 (1993). ArticleADSGoogle Scholar
81. Virtanen, P. et al. SciPy 1.0: fundamental algorithms for scientific computing in Python.Nat. Methods17, 261–272 (2020). ArticlePubMedPubMed CentralGoogle Scholar
82. Hunter, K. et al. Data for ‘Rational design of disordered proteins for sequence–function investigation’.Zenodohttps://doi.org/10.5281/zenodo.18717144(2026).

Download references

Acknowledgements

We thank the members of the Water and Life Interface Institute (WALII) for discussions; M. Staller, M. Suskiewicz, S. Dutcher, T. Lohman, G. Amarasinghe, R. Pappu, M. Kjaergaard, S. Plassmeyer, J. Pelham, G. Ginell and the staff at FNZ for many discussions and comments on the manuscript. We thank S. Minhas for the original GOOSE logo.

Funding

This work was funded by a US National Science Foundation (NSF) IntBio Collaborative Research Proposal to S.S. (2128067) and A.S.H. (2128068), a NSF BII Award (WALII, NSF DBI grant 2213983) to S.S. and A.S.H., a NSF CAREER award (2338129) to A.S.H., a Research Grant from the Human Frontiers in Science Program to A.S.H. (RGP0015/2022) and by the US NIH under award R35GM137926 to S.S. and DP2-CA290639 to A.S.H. S.S. acknowledges support from the Alfred P. Sloan Foundation. K.G. was supported by a fellowship from NSF-CREST Center for CCBM at UC Merced, grant no. NSF-HRD-1547848. K.H. was supported by the US NIH under award number T32GM141862. T.B. was supported in part by funds from the Syracuse University School of Arts and Sciences. J.M.L. was supported by an NSF GRFP (DGE-2139839).

Author information

Author notes

1. These authors contributed equally: Kara Hunter, Trevor Brandt, Karina Guadalupe

Authors and Affiliations

1. Department of Chemistry, Syracuse University, Syracuse, NY, USA Kara Hunter, Trevor Brandt, Kavindu C. Kolamunna, Jessica K. Niblo, Brooke Nicholson, Lea M. Day & Shahar Sukenik
2. Department of Chemistry and Biochemistry, University of California, Merced, Merced, CA, USA Kara Hunter, Karina Guadalupe, Nora M. Shamoon, Alec Martinez & Shahar Sukenik
3. Department of Biochemistry and Molecular Biophysics, Washington University School of Medicine, St Louis, MO, USA Jeffrey M. Lotthammer, Alex S. Holehouse & Ryan J. Emenecker
4. Center for Biomolecular Condensates (CBC), Washington University in St Louis, St Louis, MO, USA Jeffrey M. Lotthammer, Alex S. Holehouse & Ryan J. Emenecker

Authors

1. Kara Hunter
2. Trevor Brandt
3. Karina Guadalupe
4. Kavindu C. Kolamunna
5. Jeffrey M. Lotthammer
6. Nora M. Shamoon
7. Jessica K. Niblo
8. Brooke Nicholson
9. Lea M. Day
10. Alec Martinez
11. Alex S. Holehouse
12. Shahar Sukenik
13. Ryan J. Emenecker

Contributions

R.J.E., A.S.H. and S.S. conceptualized the study. R.J.E., A.S.H. and S.S. designed the methodology. R.J.E., A.S.H. and J.M.L. contributed to software development. K.H., T.B., K.G., R.J.E., N.M.S., B.N., K.C.K., J.K.N. and L.M.D. generated experimental data. T.B., K.G., N.M.S. and A.M. conducted in-cell FRET experiments. K.C.K. performed in vitro FRET experiments. J.K.N. conducted molecular dynamics simulations. R.J.E. performed yeast self-assembly experiments. K.H., B.N. and L.M.D. generated data for the yeast desiccation library. A.S.H., S.S., R.J.E., T.B., K.G. and K.H. carried out data analysis. A.S.H., S.S., J.M.L. and K.H. provided resources and acquired funding. K.H., T.B., K.G., A.S.H., S.S. and R.J.E. performed visualization and wrote the manuscript. R.J.E. formatted the Extended Data and Supplementary Information.

Corresponding authors

Correspondence toAlex S. Holehouse,Shahar SukenikorRyan J. Emenecker.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Naturethanks the anonymous reviewers for their contribution to the peer review of this work.Peer reviewer reportsare available.

Additional information

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

Extended data figures and tables

Extended Data Fig. 1 GOOSE can rapidly generate sequences and sample sequence space outside of naturally occurring IDRs.

a. Uniform Manifold Projection (UMAP) showing projected compositional space occupied by 150-amino acid fragments from the human proteome (left) or generated by GOOSE (right). UMAP was generated by embedding sequences into an L = 20 vector, where each element corresponds to the fraction of each amino acid. Human proteome fragments were identified by segmenting the human proteome into disordered regions and folded domains, then taking 150-amino-acid subfragments within those regions (step size of 50). This yielded 33,626 IDR fragments and 85,098 folded fragments. For GOOSE, we generated a minimum of 10 sequences by titrating across amino acid fractions, properties (NCPR, FCR, hydropathy, kappa), and amino acid classes (aromatic, aliphatic, polar, positive, negative, glycine, proline, cysteine, histidine), titrating from minimum to maximum possible values (step size 0.1). This yielded 67,427 150-residue GOOSE-derived segments.b. Time to generate sequences where only a single property is specified. Times are averaged over 200 generated sequences.c. Time to generate sequences where four properties are specified. Each point represents the time required to generate a single sequence. Each plot shows the property for that sequence. A total of 1,656 generated sequences are shown as they titrate across hydrophobicity, FCR, NCPR, and kappa values.d. Time to generate sequences where different numbers of amino acid fractions are specified. A total of 100 sequences were generated for each number of fractions specified. Each point is an individually generated sequence. The orange ‘X’ shows the average value across the 100 sequences. Forb.-d., all sequences are 200 amino acids in length.

Extended Data Fig. 2 Library variants described in Figs.2and3.

Columns reported here include the Fraction of Charged Residues (FCR), Net Charge Per Residue (NCPR), charge-distribution metric kappa (κ), Kyte-Doolittle hydrophobicity (H), average disorder prediction (Disorder), and a self-interaction (self ε - lower means on average stronger self-interactions). Note that a sequence must have both positive and negative residues for κ to be calculated; otherwise, it is reported as -1. Disorder here is calculated as the average disorder score as predicted by metapredict v2.0. Self-interaction score (self ε) is calculated using the CALVADOS version of the FINCHES model.

Extended Data Fig. 3 Ensemble differs between in-cell and in silico analyses when charges are clustered.

(a) Pairwise comparisons of in silico (blue) Reand in cell (green) FRET efficiencies. In silico points are the mean end-to-end distance (Ree) of 3 converged simulation repeats, and error bars are their standard deviation. In cell markers and error bars are the average and standard deviation of the medians of all biological repeats (n = listed in Fig.2a). Significance is determined by an unpaired two-sided Student’s t-test without multiple testing correction, where the asterisks associated with each sequence pair denote the resulting p-values: * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001. Sequences with evenly distributed net-positive charges all increase in ensemble dimension as the fraction of charged residues increases. (b) For the same sequence parameters but with clustered charges, in silico sequences display no significant ensemble change, while in cell sequences expand. This may imply environmental conditions influencing ensemble behaviour in cells. (c) Schematic of how GOOSE was used to generate sequences tested. All sequences here are 60 residues. Five sequences are generated with specified parameters; Reis predicted using ALBATROSS for each sequence, and the mean is reported as the condition average. This is repeated, titrating along both the number of prolines and charged residues, until the heatmaps in panels d and e are complete. (d) Heatmap of average Reas proline content and the number of charged residues increase, with evenly distributed charges. Here, as both the number of prolines and the number of charged residues increase, Reincreases. Compared sequences for which proline content differs in Fig.2fare marked as labelled stars. (e) Heatmap for sequences with clustered charges. Here, there is no consistent dependence on the number of charged residues, but increases in proline content expand the ensemble. (f) Same format as panels a and b, but with predicted Reinstead of simulated Re. In cell Efmarkers and error bars remain the average and standard deviation of the medians of all biological repeats (n = listed in Fig.2a). Predicted Remarkers are the mean of the condition average, while the error bars are the standard deviation (n = 5 sequences). Significance is determined by an unpaired two-sided Student’s t-test without multiple testing correction, where the asterisks associated with each sequence pair denote the resulting p-values: * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001. For evenly distributed charges, all ensembles expand as FCR increases and/or as proline counts increase. (g) When charges are clustered, increasing FCR and proline count concomitantly (left) expands the ensemble under both predicted and in-cell conditions, whereas when only FCR is significantly increased (right), the ensemble compacts in cell but shows no predicted change.

Extended Data Fig. 4 In vitro vs. in-cell comparison of disordered protein ensembles.

Pairwise Efcomparisons of in vitro (purple) and in-cell (green) data. In vitro data points are the mean of biological replicates (n = 3), and error bars are their standard deviation. In-cell markers and error bars are the average and standard deviation of the medians of all biological repeats (n = listed in Fig.2a). Significance is determined by an unpaired two-sided Welch’s t-test without multiple testing correction. Asterisks below each pair denote resulting p-values: * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001. (a) Effor sequences with evenly distributed charges and a constant net charge of −0.3. (b) Effor sequences that introduce aromatics and modulate hydrophobicity. In both comparisons, ensemble compaction is expected: overall, removing Q/N promotes compaction because of the loss of hydrogen-bonding partners. The addition of hydrophobic aromatics (W) that also engage in self-attractive pi-pi and cation-pi interactions strengthens this effect. Mutations to relatively more hydrophobic residues, S and V, would also be expected to contribute to ensemble compaction. (c) Efcomparisons for sequences that modulate FCR, net charge, and charge clustering. In-cell structural ensembles compact monotonically when positive charges are added, with charge patterning and FCR having little effect on this relationship. In vitro constructs behave in a more variable, sequence-dependent manner.

Extended Data Fig. 5 Images of yeast expressing scaffold constructs, client constructs, or scaffold and client constructs together.

a. mTagBFP2-Scaffold1 andb. mScarlet3-Scaffold2 alone.c. Image of yeast expressing client construct alone (furthest left) or client construct with scaffold (right three columns). The scaffold construct was mTagBFP2-Scaffold1. All clients were 200 amino acids and used a mScarlet3 fluorophore.d. Image of yeast expressing client construct alone (furthest left) or client construct with scaffold (right three columns). The scaffold construct was mScarlet3-Scaffold2. All clients were 400 amino acids in length and used the mTagBFP2 fluorophore. For all images, the scale bar is 5 µm.

Extended Data Fig. 6 Template sequences for protectant library reported in Fig.5.

Columns reported here include the UniProt identifier, if applicable, for each template sequence, the origin (if synthetically generated, if a LEA, and if it was a tile of a longer sequence), the organism it is found in, and references if studied or annotated.

Extended Data Fig. 7 Protection differs across templates, and the library confers faster regrowth post desiccation.

a. Violin plots of all templates and their corresponding weighted log2FoldChange values across 3 separate repeats of the experiment. The wider the violin, the more frequently that specific log2FoldChange value occurs. Visualized were the WT template value, the IQR, and the median (white circle).b. Plotted cell concentrations of libraries (blue, yellow, red) and an empty vector control (green) over time, measured by optical density in Pioreactors. The optical density was calibrated and converted to cell concentration (cells/mL) by correlating OD with cell counts (DeNovix Celldrop FLi). The Pioreactor was programmed to remove 2 mL of cells and add 2 mL of fresh media to dilute the culture when the cell concentration reached 2.0 × 107cells/mL. This ensures the culture remains in log phase. Removal and addition of fresh media is visualized in library 3 (red) at 15 h, and in the empty vector (EV) control at 25 h.

Extended Data Fig. 8 Comparative distribution of top and bottom-performing sequence variants.

Violin plots of the worst-performing sequences (‘bottom’, purple; log2(fold change) ≤ −0.5, statistically significant) and top-performing sequences (‘top’, green; log2(fold change) ≥ 0.5, statistically significant). N = # labelled below each category. Individual sequences are shown as black points. The middle line within each violin indicates the median; top and bottom marks indicate the maximum and minimum values. Significance between distributions of top and bottom groups was determined by a two-tailed Mann-Whitney U test, corrected for multiple testing using the Benjamini-Hochberg procedure to control the false discovery rate (FDR), and significant results (p < 0.05) are highlighted with a red border. Exact p-values are denoted above each plot. Sequence length is displayed as a boxplot; the horizontal line indicates the median, box edges indicate the interquartile range (IQR), whiskers extend to 1.5 × IQR, and individual points beyond this range are shown as outliers. All other sequence features are described in Supplementary Table4.

Extended Data Fig. 9 Comparison of sequence features to the yeast IDRome; distribution and significance of log2-fold change across sequence variants.

a. Histograms of sequence features for the yeast IDRome (grey, n = 5,555) and the top-performing library sequences (green, n = 689; inverse-variance weighted log2(fold change) > 0, BH-adjusted p-value < 0.05). The yeast IDRome distribution is shown as a kernel density estimate (KDE) scaled to sequence counts using Scott’s bandwidth rule, plotted on the right y-axis; top performer counts are shown as a step histogram on the left y-axis. Statistical significance between distributions was determined using a two-sided Mann-Whitney U test with Benjamini-Hochberg multiple testing correction; exact p-values are shown in the upper right of each panel for comparisons with a rank-biserial correlation |r| ≥ 0.1. Inverted triangles mark individual known desiccation-protective sequences: CAHS−106094, yeast HSP12, LEA14 (O03983), and LEA14-like (O82355). Sequence features were calculated as in Supplementary Table4.b. Volcano plots of the weighted log2(fold change) and the -log10(BH-adjusted p-value), as calculated in Fig.5c, highlight the original template sequence (black square) and its respective variants generated by GOOSE.

Extended Data Fig. 10 Cellular localization and puncta formation of GOOSE-generated sequences.

Representative images of the direct acceptor channel from all 40 constructs generated by GOOSE were used for Efmeasurements. Numbers in the top-left of each box correspond to sequence numbers from Extended Data Fig.2and Supplementary Table1. Blue boxes around sequences indicate sequences with two or more patches of three or more positive amino acids. Almost all constructs with runs of positively charged residues (excluding #27, 9/10) display nuclear puncta consistent with inclusion into RNA-ribonucleoprotein bodies, as well as strong nuclear localization (e.g., #10). When sequences lack clustered positive charges or contain mixed positive and negative charges, they display a diffuse signal (e.g., #30). Scale bar = 50 µm.

Supplementary information

Rights and permissions

Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.

Reprints and permissions

About this article

Cite this article

Hunter, K., Brandt, T., Guadalupe, K.*et al.*Rational design of disordered proteins for sequence–function investigation.Nature(2026). https://doi.org/10.1038/s41586-026-10849-1

Download citation

• Received:27 October 2023
• Accepted:24 June 2026
• Published:29 July 2026
• Version of record:29 July 2026
• DOI:https://doi.org/10.1038/s41586-026-10849-1