A Vast Navigable Protein Sequence Subspace
Open accessAbstract
Fitness landscapes govern evolution. Our empirical understanding of their structure has primarily been informed by evolutionary history. Here, we combine high-throughput DNA synthesis and machine learning, without training on evolutionary priors, to build a high-resolution map of the genome packaging landscape of the Adeno-associated Virus capsid protein. We assay 232,123 mutants around Adeno-Associated Virus 2 (AAV2) including 15 neighborhoods centered on seed sequences within 5-21 mutations of the wild type sequence. Local neighborhoods differ sharply in mutational tolerance and epistatic detail, yet much of the quantitative variation is captured by a simple pattern: sitewise effects transformed by global epistasis. Strikingly, using targeted sampling of only a few thousand mutants, we uncover a vast percolating network of function-preserving ridges embedded within a largely non-functional landscape. These ridges support billions of distinct trajectories of up to 667 steps that connect all sampled neighborhoods through functional single mutations. Designed variants with mutations distinct from the evolutionary background retain packaging at distances where randomly generated mutants rarely survive. These include capsids that combine up to 7 mutations that are largely absent from the evolutionary record, and which are each individually lethal in the 15 diverse backgrounds we tested. Our results reveal rich subspace structure resembling landscape orebodies: reservoirs of highly viable and navigable sequence space defined by simple constraints. These subspaces can be discovered through targeted sampling and can open access to vast troves of proteins that may be valuable for downstream applications such as therapeutics.
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.
Integrity signals
Registry
Cites
Statistics
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.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Cites
Darwinian Evolution Can Follow Only Very Few Mutational Paths to Fitter Proteins
Weinreich, D. M., Delaney, N. F., DePristo, M. A., & Hartl, D. L. (2006) · Science · doi:10.1126/science.1123539 · Cited by 1,418 · see citing works
Large language models generate functional protein sequences across diverse families
Madani, A., Krause, B., Greene, E. R., Subramanian, S., Mohr, B. P., Holton, J. M., Olmos, J. L., Xiong, C., Sun, Z. Z., Socher, R., et al. (2023) · Nature Biotechnology · doi:10.1038/s41587-022-01618-2 · Cited by 1,078 · cited 1× · see citing works
“…interpretability of simple models and the precise prediction ability and high flexibility that emerges from training high-capacity models on large datasets (Puszkarska et al. 2024; Hopf et al. 2017; Hsu et al. 2022; Madani et al. 2023; D. Ding et al. 2024; Candido et al. 2026; Park et al. 2024).”
Mutation effects predicted from sequence co-variation
Hopf, T. A., Ingraham, J. B., Poelwijk, F. J., Schärfe, C. P. I., Springer, M., Sander, C., & Marks, D. S. (2017) · Nature Biotechnology · doi:10.1038/nbt.3769 · Cited by 775 · see citing works
The NK model of rugged fitness landscapes and its application to maturation of the immune response
Kauffman, S. A., & Weinberger, E. D. (1989) · Journal of Theoretical Biology · doi:10.1016/s0022-5193(89)80019-0 · Cited by 674 · see citing works
Local fitness landscape of the green fluorescent protein
Sarkisyan, K. S., Bolotin, D. A., Meer, M. V., Usmanova, D. R., Mishin, A. S., Sharonov, G. V., Ivankov, D. N., Bozhanova, N. G., Baranov, M. S., Soylemez, O., et al. (2016) · Nature · doi:10.1038/nature17995 · Cited by 591 · see citing works
Machine learning-assisted directed protein evolution with combinatorial libraries
Wu, Z., Kan, S. B. J., Lewis, R. D., Wittmann, B. J., & Arnold, F. H. (2019) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.1901979116 · Cited by 580 · see citing works
Neutral evolution of mutational robustness
van Nimwegen, E., Crutchfield, J. P., & Huynen, M. A. (1999) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.96.17.9716 · Cited by 568 · see citing works
Tabebordbar, M., Lagerborg, K. A., Stanton, A. C., King, E. M., Ye, S. H., Tellez, L., Krunnfusz, A., Tavakoli, S., Widrick, J. J., Messemer, K. A., et al. (2021) · Cell · doi:10.1016/j.cell.2021.08.028 · Cited by 517 · see citing works
Low-N protein engineering with data-efficient deep learning
Biswas, S., Khimulya, G., Alley, E. C., Esvelt, K. M., & Church, G. M. (2021) · Nature Methods · doi:10.1038/s41592-021-01100-y · Cited by 422 · see citing works
Self-citationComprehensive AAV capsid fitness landscape reveals a viral gene and enables machine-guided design
Ogden, P. J., Kelsic, E. D., Sinai, S., & Church, G. M. (2019) · Science · doi:10.1126/science.aaw2900 · Cited by 356 · see citing works
Self-citationNavigating the protein fitness landscape with Gaussian processes
Romero, P. A., Krause, A., & Arnold, F. H. (2012) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.1215251110 · Cited by 346 · see citing works
An evolution-based model for designing chorismate mutase enzymes
Russ, W. P., Figliuzzi, M., Stocker, C., Barrat-Charlaix, P., Socolich, M., Kast, P., Hilvert, D., Monasson, R., Cocco, S., Weigt, M., et al. (2020) · Science · doi:10.1126/science.aba3304 · Cited by 313 · see citing works
Deep diversification of an AAV capsid protein by machine learning
Bryant, D., Bashir, A., Sinai, S., Jain, N. K., Ogden, P. J., Riley, P. F., Church, G. M., Colwell, L. J., & Kelsic, E. D. (2021) · Nature Biotechnology · doi:10.1038/s41587-020-00793-4 · Cited by 302 · see citing works
Self-citationNeutralism and selectionism: a network-based reconciliation
Wagner, A. (2008) · Nature Reviews Genetics · doi:10.1038/nrg2473 · Cited by 302 · see citing works
Should evolutionary geneticists worry about higher-order epistasis?
Weinreich, D. M., Lan, Y., Wylie, C. S., & Heckendorn, R. B. (2013) · Current Opinion in Genetics & Development · doi:10.1016/j.gde.2013.10.007 · Cited by 277 · see citing works
Rapid in silico directed evolution by a protein language model with EVOLVEpro.
Jiang, K.-Y., Yan, Z.-Q., Bernardo, M. D., Sgrizzi, S. R., Villiger, L., Kayabolen, A., Kim, B. J., Carscadden, J. K., Hiraizumi, M., Nishimasu, H., et al. (2024) · Science · doi:10.1126/science.adr6006 · Cited by 248 · see citing works
Inferring the shape of global epistasis
Otwinowski, J., McCandlish, D. M., & Plotkin, J. B. (2018) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.1804015115 · Cited by 243 · see citing works
Learning protein fitness models from evolutionary and assay-labeled data
Hsu, C., Nisonoff, H. M., Fannjiang, C., & Listgarten, J. (2022) · Nature Biotechnology · doi:10.1038/s41587-021-01146-5 · Cited by 222 · see citing works
An AAV capsid reprogrammed to bind human transferrin receptor mediates brain-wide gene delivery
Huang, Q., Chan, K. Y., Wu, J. H., Botticello-Romero, N. R., Zheng, Q., Lou, S., Keyes, C., Svanbergsson, A., Johnston, J., Mills, A., et al. (2024) · Science · doi:10.1126/science.adm8386 · Cited by 200 · see citing works
Learning the pattern of epistasis linking genotype and phenotype in a protein
Poelwijk, F. J., Socolich, M., & Ranganathan, R. (2019) · Nature Communications · doi:10.1038/s41467-019-12130-8 · Cited by 165 · see citing works
Neural networks to learn protein sequence–function relationships from deep mutational scanning data
Gelman, S., Fahlberg, S. A., Heinzelman, P., Romero, P. A., & Gitter, A. (2021) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2104878118 · Cited by 160 · see citing works
Detecting High-Order Epistasis in Nonlinear Genotype-Phenotype Maps
Sailer, Z. R., & Harms, M. J. (2017) · Genetics · doi:10.1534/genetics.116.195214 · Cited by 154 · see citing works
Adachi, K., Enoki, T., Kawano, Y., Veraz, M., & Nakai, H. (2014) · Nature Communications · doi:10.1038/ncomms4075 · Cited by 149 · see citing works
On the (un)predictability of a large intragenic fitness landscape
Bank, C., Matuszewski, S., Hietpas, R. T., & Jensen, J. D. (2016) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.1612676113 · Cited by 124 · see citing works
A rugged yet easily navigable fitness landscape
Papkou, A., García-Pastor, L., Escudero, J. A., & Wagner, A. (2023) · Science · doi:10.1126/science.adh3860 · Cited by 122 · see citing works
Global epistasis emerges from a generic model of a complex trait
Reddy, G., & Desai, M. M. (2021) · eLife · doi:10.7554/elife.64740 · Cited by 111 · see citing works
Idiosyncratic epistasis leads to global fitness–correlated trends
Bakerlee, C. W., Ba, A. N. N., Shulgina, Y., Echenique, J. I. R., & Desai, M. M. (2022) · Science · doi:10.1126/science.abm4774 · Cited by 99 · see citing works
The structure of genotype-phenotype maps makes fitness landscapes navigable
Greenbury, S. F., Louis, A. A., & Ahnert, S. E. (2022) · Nature Ecology & Evolution · doi:10.1038/s41559-022-01867-z · Cited by 76 · see citing works
The simplicity of protein sequence-function relationships
Park, Y., Metzger, B. P. H., & Thornton, J. W. (2024) · Nature Communications · doi:10.1038/s41467-024-51895-5 · Cited by 69 · see citing works
Heterogeneity of the GFP fitness landscape and data-driven protein design
Somermeyer, L. G., Fleiss, A., Mishin, A. S., Bozhanova, N. G., Igolkina, A. A., Meiler, J., Pujol, M.-E. A., Putintseva, E. V., Sarkisyan, K. S., & Kondrashov, F. A. (2022) · eLife · doi:10.7554/elife.75842 · Cited by 68 · see citing works
Language Modeling Materializes a World Model of Protein BiologyPreprint
Candido, S., Hayes, T., Derry, A., Rao, R., Lin, Z., Verkuil, R., Wu, B. Z., Lee, J. S., Bruguera, E. S., Keval, J. A., et al. (2026) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.64898/2026.06.03.729735 · Cited by 65 · see citing works
Zhu, D., Brookes, D. H., Busia, A., Carneiro, A., Fannjiang, C., Popova, G., Shin, D., Donohue, K. C., Lin, L. F., Miller, Z. M., et al. (2024) · Science Advances · doi:10.1126/sciadv.adj3786 · Cited by 63 · see citing works
Protein language models are biased by unequal sequence sampling across the tree of life
Ding, F., & Steinhardt, J. (2024) · doi:10.1101/2024.03.07.584001 · Cited by 57 · see citing works
Neural network extrapolation to distant regions of the protein fitness landscape
Freschlin, C. R., Fahlberg, S. A., Heinzelman, P., & Romero, P. A. (2024) · Nature Communications · doi:10.1038/s41467-024-50712-3 · Cited by 50 · see citing works
Higher-order epistasis and phenotypic prediction
Zhou, J., Wong, M. S., Chen, W.-C., Krainer, A. R., Kinney, J. B., & McCandlish, D. M. (2022) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2204233119 · Cited by 46 · see citing works
Co-evolution of interacting proteins through non-contacting and non-specific mutations
Ding, D., Green, A. G., Wang, B., Lite, T.-L. V., Weinstein, E. N., Marks, D. S., & Laub, M. T. (2022) · Nature Ecology & Evolution · doi:10.1038/s41559-022-01688-0 · Cited by 43 · see citing works
Machine learning designs new GCGR/GLP-1R dual agonists with enhanced biological potency
Puszkarska, A. M., Taddese, B., Revell, J. D., Davies, G., Field, J., Hornigold, D. C., Buchanan, A. G., Vaughan, T. J., & Colwell, L. J. (2024) · Nature Chemistry · doi:10.1038/s41557-024-01532-x · Cited by 43 · see citing works
Self-citationOn the sparsity of fitness functions and implications for learning
Brookes, D. H., Aghazadeh, A., & Listgarten, J. (2021) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2109649118 · Cited by 43 · see citing works
Minimum epistasis interpolation for sequence-function relationships
Zhou, J., & McCandlish, D. M. (2020) · Nature Communications · doi:10.1038/s41467-020-15512-5 · Cited by 41 · see citing works
Protein design using structure-based residue preferences
(2024) · doi:10.1038/s41467-024-45621-4 · Cited by 38 · see citing works
Self-citationGenerative AAV capsid diversification by latent interpolationPreprint
Sinai, S., Jain, N. K., Church, G. M., & Kelsic, E. D. (2021) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2021.04.16.440236 · Cited by 34 · see citing works
Self-citationRotrattanadumrong, R., & Yokobayashi, Y. (2022) · Nature Communications · doi:10.1038/s41467-022-32538-z · Cited by 33 · see citing works
Patton, A. H., Richards, E. J., Gould, K. J., Buie, L. K., & Martin, C. H. (2022) · eLife · doi:10.7554/elife.72905 · Cited by 26 · see citing works
Regeneration of actin filament branches from the same Arp2/3 complex
Ghasemi, F., Cao, L., Mladenov, M., Guichard, B., Way, M., Jegou, A. G., & Romet‐Lemonne, G. (2024) · Science Advances · doi:10.1126/sciadv.adj7681 · Cited by 22 · see citing works
Recombination and mutational robustness in neutral fitness landscapes
Klug, A., Park, S., & Krug, J. (2019) · PLoS Computational Biology · doi:10.1371/journal.pcbi.1006884 · Cited by 14 · see citing works
Virus Evolution on Fitness Landscapes
Schuster, P. M., & Stadler, P. F. (2023) · Current topics in microbiology and immunology · doi:10.1007/978-3-031-15640-3_1 · Cited by 9 · see citing works
Contrastive losses as generalized models of global epistasis
Brookes, D. H., Otwinowski, J., & Sinai, S. (2024) · neural information processing systems · doi:10.52202/079017-2962 · Cited by 3 · see citing works
Self-citationEvolutionary paths that link orthogonal pairs of binding proteins
Avizemer, Z., Martí‐Gómez, C., Hoch, S. Y., McCandlish, D. M., & Fleishman, S. J. (2025) · Cell Systems · doi:10.1016/j.cels.2025.101262 · Cited by 2 · see citing works
Understanding protein function with a multimodal retrieval-augmented foundation modelPreprint
Truong, T. F., & Bepler, T. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2508.04724 · Cited by 2 · see citing works
Random Fitness Landscapes are Highly Navigable
Oros, D., & Krug, J. (2026) · doi:10.64898/2026.05.22.727193 · Cited by 2
Mount Fuji’s stubby peak: the genotypic density of additive landscapes near maximal fitness
Kinney, J. B. (2026) · Genetics · doi:10.1093/genetics/iyag133 · Cited by 1 · see citing works
Random fitness landscapes are highly navigable
Oros, D., & Krug, J. (2026) · Genetics · doi:10.1093/genetics/iyag251 · Cited by 1 · see citing works
Unobserved Sequence Space Has Many Functional ProteinsPreprint
Diplock, N., Desautels, T., Haque, R., Finney, M., Wackelin, D. J., Kang-Yun, C. S., Park, D. M., & Leonard, S. P. (2026) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.64898/2026.08.24.746828
Wright, Sewall , and Others. 1932. The Roles of Mutation, Inbreeding, Crossbreeding, and Selection in Evolution. http://www.esp.org/books/6th-congress/facsimile/contents/6th-cong-p356-wright.pdf.
(n.d.)
Unresolved
Flagged first, then by citations
1 of the 51 references couldn’t be matched to a known paper.
Cited by
Still being checked.
- Authors
- Sinai, S., Gane, A., Jain, N., Bryant, D., Bileschi, M., Bashir, A., Church, G. M., Kelsic, E., & Colwell, L.
- Affiliations
- Wyss Institute for Biologically Inspired Engineering (US)
- Google (United States) (US)
- Harvard University (US)
- Google Research
- Corresponding author
- Sinai, Sam
- Venue
- bioRxiv (Cold Spring Harbor Laboratory)
- Publisher
- openRxiv
- Published
- 2026-10-22
- OpenAlex
- W7223249742
- Cited by
- 0
- References
- 51
- Type
- posted content
- Language
- en
- Open access
- green
- doi.org
- Licence
- Locations
- bioRxiv (Cold Spring Harbor Laboratory) (acceptedVersion, cc-by)
- Funders
- Hansjörg Wyss Institute for Biologically Inspired Engineering, Harvard University
- Topics
- Life Sciences › Biochemistry, Genetics and Molecular Biology › Genetics › Evolution and Genetic Dynamics
- Life Sciences › Biochemistry, Genetics and Molecular Biology › Genetics › Virus-based gene therapy research
- Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › RNA and protein synthesis mechanisms
- Keywords
- fitness landscape, genome packaging, epistasis, mutational tolerance, sequence spaces, ore bodies, protein evolution, machine learning
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
