pith. sign in

arxiv: 2606.29529 · v1 · pith:2RNOTMZKnew · submitted 2026-06-28 · 🧬 q-bio.PE

Modeling Protein Evolution with Generative Models: from Extant Sequence Data to Evolutionary Dynamics

Matteo Bisardi , Leonardo Di Bari , Saverio Rossi , Karol Buda , Francesco Zamponi , Martin Weigt This is my paper

Pith reviewed 2026-06-30 01:48 UTC · model grok-4.3

classification 🧬 q-bio.PE
keywords protein evolutiongenerative modelsdirect coupling analysissequence landscapesevolutionary dynamicsepistasispopulation geneticshomologous sequences
0
0 comments X

The pith

Generative models trained on protein sequence families define probabilistic landscapes that can be coupled to population-genetic dynamics to simulate evolutionary trajectories.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper examines how models such as Direct Coupling Analysis, trained on alignments of related protein sequences, produce scoreable landscapes that encode mutation, selection, and epistatic constraints. These landscapes are then linked to substitution or population-genetic models to generate forward simulations of sequence change on laboratory or phylogenetic timescales. Applications discussed include viral adaptation, laboratory drift, historical contingency, and long-term exploration of sequence space. A sympathetic reader would see this as a route from static sequence databases to dynamic evolutionary forecasts without requiring exhaustive fitness measurements for every variant.

Core claim

Generative sequence landscapes inferred from homologous families can be coupled to population-genetic or substitution-model dynamics to simulate protein evolution across experimental and phylogenetic timescales, with Direct Coupling Analysis serving as a validated, interpretable instance of the approach.

What carries the argument

Direct Coupling Analysis, which infers an energy function from multiple-sequence alignments to assign probabilities to entire sequences and thereby defines the landscape used for dynamics.

If this is right

  • Simulations of viral evolution under changing selective pressures become feasible from sequence data alone.
  • Laboratory drift experiments can be modeled in silico to predict reachable sequence variants.
  • Historical contingency and entrenchment effects can be quantified by replaying evolution from different ancestral states.
  • Epistatic contributions to substitution rates can be tracked over extended timescales.
  • Long-term exploration of viable sequence space can be performed without enumerating all possible mutants.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same coupling could be tested against codon-level mutation spectra to check whether the inferred landscapes remain predictive when mutation biases are included explicitly.
  • Integration with structural data might reveal whether the landscapes implicitly encode three-dimensional constraints that are not captured by sequence statistics alone.
  • If calibration between model scores and actual fitness improves, the framework could supply priors for forecasting which mutations are likely to fix in clinical or agricultural settings.
  • Extension to include indels would require new generative architectures but could address a major gap in current sequence-space models.

Load-bearing premise

Landscapes fitted to present-day sequences are assumed to encode the functional constraints that governed past evolution and will govern future evolution without major distortion from sampling or model misspecification.

What would settle it

A direct comparison in which trajectories simulated from the coupled model systematically fail to reproduce the substitution patterns or fitness changes observed in a controlled laboratory evolution experiment or a well-resolved phylogenetic clade for the same protein.

Figures

Figures reproduced from arXiv: 2606.29529 by Francesco Zamponi, Karol Buda, Leonardo di Bari, Martin Weigt, Matteo Bisardi, Saverio Rossi.

Figure 1
Figure 1. Figure 1: Schematic representation of generative sequence landscape. (a) Schematic representation of a protein fitness landscape and its corresponding sequence space. Each point on the fitness landscape represents a protein sequence, and its height represents fitness. Extant natural sequences occupy functional regions of the landscape, whereas non-functional sequences lie in low-fitness regions. Experimental approac… view at source ↗
Figure 2
Figure 2. Figure 2: Experimental validation of generative sequence landscapes across evolutionary scales. (a) Schematic representation of the different evolutionary scales over which sequence landscapes can be compared with and validated against experimental data. Increasingly divergent regions of sequence space with respect to the hypothetical wild-type sequence SANGER (white star) are illustrated by shaded areas of differen… view at source ↗
Figure 3
Figure 3. Figure 3: Population genetics and substitution models on generative landscapes. (a) Schematic representation of a population genetics cycle on a sequence landscape: mutagenesis introduces diversity, selection filters sequences based on their generative score ϕ(s), and amplification propagates functional variants. The survival probability is determined by the fitness (score) of the sequence relative to the population… view at source ↗
Figure 4
Figure 4. Figure 4: Results on experimentally accessible timescales. (a) Predicted versus reported timing of HIV-1 immune escape (left, measured in days) [116] and acquisition of drug-resistance mutations (right, measured in months) [122], demonstrating the accuracy of generative evolutionary simulations in modeling virus evolution. (b) Positive predictive power (PPV) for structural contacts of the PSE-1 gene by DCA as a func… view at source ↗
Figure 5
Figure 5. Figure 5: Results beyond currently accessible experimental timescales. (a) In the left panel, local site mutability is compared to global site mutability for the PSE-1 gene, with residues colored by category: locally and globally mutable (red), locally and globally conserved (blue), and locally conserved but globally mutable (violet). In gray, all other residues. The right panel shows the temporal evolution of site … view at source ↗
read the original abstract

Protein sequences carry a record of evolutionary history shaped by mutation, selection, drift, and epistasis. Recent generative models trained on homologous sequence families offer a new way to read this record: they define probabilistic landscapes that score sequences, generate viable variants, and capture constraints that are difficult to measure experimentally. In this review, we discuss how such landscapes can be used not only for protein design or mutation-effect prediction, but also for modeling evolutionary dynamics. We focus particularly on Direct Coupling Analysis as an interpretable and experimentally validated framework, while placing it in the broader context of generative sequence modeling. We first describe how generative sequence landscapes are inferred and assessed, then review how they can be coupled to population-genetic or substitution-model dynamics to simulate protein evolution across experimental and phylogenetic timescales. Applications include viral evolution, laboratory drift experiments, historical contingency, entrenchment, epistatic drift over time, and long-term sequence-space exploration. We conclude by discussing open challenges, including score-fitness calibration, phylogenetic structure, codon-level mutation biases, indels, and the integration of experimental data.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 1 minor

Summary. The manuscript is a review summarizing how generative models (with emphasis on Direct Coupling Analysis) trained on homologous protein sequence families infer probabilistic landscapes that capture evolutionary constraints. These landscapes are then coupled to population-genetic or substitution-model dynamics to simulate protein evolution on experimental and phylogenetic timescales. The review covers landscape inference and validation, applications (viral evolution, laboratory drift, historical contingency, entrenchment, epistatic drift), and open challenges (score-fitness calibration, phylogenetic structure, codon biases, indels, experimental data integration). No new empirical results or derivations are presented; the central contribution is descriptive synthesis of existing literature.

Significance. If the coverage is accurate and balanced, the review offers a timely synthesis bridging generative sequence modeling and evolutionary dynamics, which could help researchers navigate connections between DCA-style approaches and population-genetic simulations. Explicitly flagging unresolved issues (rather than claiming resolution) is a strength. As a review without new results, its significance rests on the quality of literature representation and the utility of the outlined framework for future work.

minor comments (1)
  1. [Abstract] Abstract: the phrase 'we first describe how generative sequence landscapes are inferred and assessed, then review how they can be coupled...' would benefit from an explicit section-by-section outline early in the introduction to improve navigation for readers.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their positive and accurate summary of the manuscript, which correctly identifies it as a review synthesizing existing literature on generative sequence models (with emphasis on DCA) for evolutionary dynamics. We appreciate the recommendation to accept and the recognition that explicitly flagging open challenges is a strength.

Circularity Check

0 steps flagged

No significant circularity in this review manuscript

full rationale

This manuscript is explicitly a review summarizing prior literature on generative sequence models (primarily DCA) and their coupling to population-genetic or substitution dynamics. It contains no new derivations, equations, fitted parameters, or quantitative predictions. All central claims are descriptive of approaches already explored in the cited external works; open challenges are flagged without resolution. No load-bearing step reduces by construction to inputs or self-citations.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

This is a review paper with no new mathematical models, empirical claims, or derivations; therefore it introduces no free parameters, axioms, or invented entities of its own.

pith-pipeline@v0.9.1-grok · 5735 in / 1063 out tokens · 33521 ms · 2026-06-30T01:48:53.742863+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

169 extracted references · 28 canonical work pages · 4 internal anchors

  1. [1]

    E. L. Van Dijk, H. Auger, Y. Jaszczyszyn, and C. Ther- mes, Ten years of next-generation sequencing technol- ogy, Trends in genetics30, 418 (2014)

    2014

  2. [2]

    D. H. Ackley, G. E. Hinton, and T. J. Sejnowski, A learning algorithm for boltzmann machines, Cognitive science9, 147 (1985)

    1985

  3. [3]

    Vaswani, N

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin, At- tention is all you need, Advances in neural information processing systems30, 10. 48550/ arXiv. 1706. 03762 (2017)

    2017

  4. [4]

    Z. Lin, H. Akin, R. Rao, B. Hie, Z. Zhu, W. Lu, N. Smetanin, R. Verkuil, O. Kabeli, Y. Shmueli,et al., Evolutionary-scale prediction of atomic-level protein structure with a language model, Science379, 1123 (2023)

    2023

  5. [5]

    Stärk, C

    H. Stärk, C. Dallago, M. Heinzinger, and B. Rost, Light attention predicts protein location from the language of life, Bioinformatics Advances1, vbab035 (2021)

    2021

  6. [6]

    Cheng, C

    P. Cheng, C. Mao, J. Tang, S. Yang, Y. Cheng, W. Wang, Q. Gu, W. Han, H. Chen, S. Li,et al., Zero- shot prediction of mutation effects with multimodal deep representation learning guides protein engineering, Cell Research34, 630 (2024)

    2024

  7. [7]

    Madani, B

    A. Madani, B. Krause, E. R. Greene, S. Subramanian, B. P. Mohr, J. M. Holton, J. L. Olmos, C. Xiong, Z. Z. Sun, R. Socher, J. S. Fraser, and N. Naik, Large language models generate functional protein sequences across diverse families, Nature Biotechnology41, 1099 (2023)

    2023

  8. [8]

    Fragata, A

    I. Fragata, A. Blanckaert, M. A. D. Louro, D. A. Liber- les, and C. Bank, Evolution in the light of fitness land- scape theory, Trends in ecology & evolution34, 69 (2019)

    2019

  9. [9]

    J. F. Kingman, A simple model for the balance between selection and mutation, Journal of Applied Probability 15, 1 (1978)

    1978

  10. [10]

    Kauffman and S

    S. Kauffman and S. Levin, Towards a general theory of adaptive walks on rugged landscapes, Journal of theo- retical Biology128, 11 (1987)

    1987

  11. [11]

    S. A. Kauffman and E. D. Weinberger, The nk model of rugged fitness landscapes and its application to mat- uration of the immune response, Journal of theoretical biology141, 211 (1989)

    1989

  12. [12]

    R. A. Fisher,The genetical theory of natural selection: a complete variorum edition(Oxford University Press, 1999)

    1999

  13. [13]

    Tenaillon, The utility of fisher’s geometric model in evolutionary genetics, Annual review of ecology, evolu- tion, and systematics45, 179 (2014)

    O. Tenaillon, The utility of fisher’s geometric model in evolutionary genetics, Annual review of ecology, evolu- tion, and systematics45, 179 (2014)

    2014

  14. [14]

    Pahujani and J

    S. Pahujani and J. Krug, Complexity and accessibility of random landscapes, SciPost Phys. Lect. Notes , 108 (2025)

    2025

  15. [15]

    Evolution as fitness landscape navigation: Concepts, Measures, and Emerging Questions

    M. Srivastava, C. Bank, J. Krug, and S. G. Das, Evo- lution as fitness landscape navigation: Concepts, mea- sures, and emerging questions, arXiv 2604.17036 (2026)

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  16. [16]

    Simple sign epistasis and evolutionary detours in fitness landscapes

    P. Ribeca, A. Castro, A. Lage-Castellanos, A. Sergeeva, S. Matuszewski, R. G. Paccosi, V. Belik, M. Ghafari, J. Krug, G. Achaz, and L. Ferretti, Simple sign epistasis and evolutionary detours in fitness landscapes, arXiv 2604.22611 (2026)

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  17. [17]

    D. M. Fowler and S. Fields, Deep mutational scanning: a new style of protein science, Nature methods11, 801 (2014)

    2014

  18. [18]

    Firnberg, J

    E. Firnberg, J. W. Labonte, J. J. Gray, and M. Oster- meier, A comprehensive, high-resolution map of a gene’s fitness landscape, Molecular biology and evolution31, 1581 (2014)

    2014

  19. [19]

    F. J. Poelwijk, M. Socolich, and R. Ranganathan, Learning the pattern of epistasis linking genotype and phenotype in a protein, Nature communications10, 4213 (2019)

    2019

  20. [20]

    Papkou, L

    A. Papkou, L. Garcia-Pastor, J. A. Escudero, and A. Wagner, A rugged yet easily navigable fitness land- scape, Science382, eadh3860 (2023)

    2023

  21. [21]

    K. Buda, C. M. Miton, and N. Tokuriki, Pervasive epistasis exposes intramolecular networks in adaptive enzyme evolution, Nature Communications14, 8508 (2023)

    2023

  22. [22]

    J. A. G. De Visser and J. Krug, Empirical fitness land- scapes and the predictability of evolution, Nature Re- views Genetics15, 480 (2014)

    2014

  23. [23]

    T. N. Starr and J. W. Thornton, Epistasis in protein evolution, Protein science25, 1204 (2016)

    2016

  24. [24]

    M. S. Johnson, G. Reddy, and M. M. Desai, Epistasis and evolution: recent advances and an outlook for pre- diction, BMC biology21, 120 (2023)

    2023

  25. [25]

    C. M. Miton and N. Tokuriki, How mutational epistasis impairs predictability in protein evolution and design, Protein Science25, 1260 (2016)

    2016

  26. [26]

    M. Blum, A. Andreeva, L. C. Florentino, S. R. Chugu- ransky, T. Grego, E. Hobbs, B. L. Pinto, A. Orr, T. Paysan-Lafosse, I. Ponamareva,et al., Interpro: the protein sequence classification resource in 2025, Nucleic Acids Research , gkae1082 (2024)

    2025

  27. [27]

    D. A. Liberles, S. A. Teichmann, I. Bahar, U. Bas- tolla, J. Bloom, E. Bornberg-Bauer, L. J. Colwell, A. J. De Koning, N. V. Dokholyan, J. Echave,et al., The interface of protein structure, protein biophysics, and molecular evolution, Protein Science21, 769 (2012)

    2012

  28. [28]

    Sikosek and H

    T. Sikosek and H. S. Chan, Biophysics of protein evo- lution and evolutionary protein biophysics, Journal of The Royal Society Interface11, 10.1098/rsif.2014.0419 (2014)

    work page doi:10.1098/rsif.2014.0419 2014

  29. [29]

    Bastolla, Y

    U. Bastolla, Y. Dehouck, and J. Echave, What evolu- tion tells us about protein physics, and protein physics tells us about evolution, Current opinion in structural biology42, 59 (2017)

    2017

  30. [30]

    Echave and C

    J. Echave and C. O. Wilke, Biophysical models of pro- tein evolution: understanding the patterns of evolution- ary sequence divergence, Annual review of biophysics 46, 85 (2017)

    2017

  31. [31]

    A. S. Lapedes, B. G. Giraud, L. Liu, and G. D. Stormo, Correlated mutations in models of protein se- quences: phylogenetic and structural effects, Lecture Notes-Monograph Series , 236 (1999)

    1999

  32. [32]

    Weigt, R

    M. Weigt, R. A. White, H. Szurmant, J. A. Hoch, and T. Hwa, Identification of direct residue contacts in protein–protein interaction by message passing, Pro- ceedings of the National Academy of Sciences106, 67 (2009). 23

    2009

  33. [33]

    Bjerregaard, P

    A. Bjerregaard, P. M. Groth, S. Hauberg, A. Krogh, and W. Boomsma, Foundation models of protein sequences: A brief overview, Current Opinion in Structural Biology 91, 103004 (2025)

    2025

  34. [34]

    Figliuzzi, P

    M. Figliuzzi, P. Barrat-Charlaix, and M. Weigt, How pairwise coevolutionary models capture the collective residue variability in proteins?, Molecular biology and evolution35, 1018 (2018)

    2018

  35. [35]

    McGee, S

    F. McGee, S. Hauri, Q. Novinger, S. Vucetic, R. M. Levy, V. Carnevale, and A. Haldane, The generative ca- pacity of probabilistic protein sequence models, Nature communications12, 6302 (2021)

    2021

  36. [36]

    Shekhar, C

    K. Shekhar, C. F. Ruberman, A. L. Ferguson, J. P. Bar- ton, M. Kardar, and A. K. Chakraborty, Spin models inferred from patient-derived viral sequence data faith- fully describe hiv fitness landscapes, Physical Review E—Statistical, Nonlinear, and Soft Matter Physics88, 062705 (2013)

    2013

  37. [37]

    Kaltenbach and N

    M. Kaltenbach and N. Tokuriki, Dynamics and con- straints of enzyme evolution, Journal of Experimental Zoology Part B: Molecular and Developmental Evolu- tion322, 468 (2014)

    2014

  38. [38]

    Morcos, T

    F. Morcos, T. Hwa, J. N. Onuchic, and M. Weigt, Di- rect coupling analysis for protein contact prediction, in Protein structure prediction(Springer, 2014) pp. 55–70

    2014

  39. [39]

    R. M. Levy, A. Haldane, and W. F. Flynn, Potts hamil- tonian models of protein co-variation, free energy land- scapes, and evolutionary fitness, Current opinion in structural biology43, 55 (2017)

    2017

  40. [40]

    Cocco, C

    S. Cocco, C. Feinauer, M. Figliuzzi, R. Monasson, and M. Weigt, Inverse statistical physics of protein se- quences: a key issues review, Reports on Progress in Physics81, 032601 (2018)

    2018

  41. [41]

    Haldane and R

    A. Haldane and R. M. Levy, Influence of multiple- sequence-alignment depth on potts statistical models of protein covariation, Physical Review E99, 032405 (2019)

    2019

  42. [42]

    Tiana and R

    G. Tiana and R. Broglia, The molecular evolution of HIV-1 protease simulated at atomic detail, Proteins: Structure, Function, and Bioinformatics76, 895 (2009)

    2009

  43. [43]

    Crippa, D

    M. Crippa, D. Andreghetti, R. Capelli, and G. Tiana, Evolutionoffrustratedandstabilisingcontactsinrecon- structed ancient proteins, European Biophysics Journal 50, 699 (2021)

    2021

  44. [44]

    Socolich, S

    M. Socolich, S. W. Lockless, W. P. Russ, H. Lee, K. H. Gardner, and R. Ranganathan, Evolutionary informa- tion for specifying a protein fold, Nature437, 512 (2005)

    2005

  45. [45]

    Bisardi, J

    M. Bisardi, J. Rodriguez-Rivas, F. Zamponi, and M. Weigt, Modeling sequence-space exploration and emergence of epistatic signals in protein evolution, Molecular biology and evolution39, msab321 (2022)

    2022

  46. [46]

    Lunzer, G

    M. Lunzer, G. B. Golding, and A. M. Dean, Pervasive cryptic epistasis in molecular evolution, PLoS genetics 6, e1001162 (2010)

    2010

  47. [47]

    Biswas, A

    A. Biswas, A. Haldane, E. Arnold, and R. M. Levy, Epistasis and entrenchment of drug resistance in hiv-1 subtype b, Elife8, e50524 (2019)

    2019

  48. [48]

    Domingo, P

    J. Domingo, P. Baeza-Centurion, and B. Lehner, The causes and consequences of genetic interactions (epista- sis), Annual review of genomics and human genetics20, 433 (2019)

    2019

  49. [49]

    Y. Park, B. P. Metzger, and J. W. Thornton, Epistatic drift causes gradual decay of predictability in protein evolution, Science376, 823 (2022)

    2022

  50. [50]

    J. Chen, M. Bisardi, D. Lee, S. Cotogno, F. Zamponi, M.Weigt,andN.Tokuriki,Understandingepistaticnet- works in the b1β-lactamases through coevolutionary statistical modeling and deep mutational scanning, Na- ture communications15, 8441 (2024)

    2024

  51. [51]

    Ekeberg, C

    M. Ekeberg, C. Lövkvist, Y. Lan, M. Weigt, and E. Au- rell, Improved contact prediction in proteins: using pseudolikelihoods to infer potts models, Physical Re- view E—Statistical, Nonlinear, and Soft Matter Physics 87, 012707 (2013)

    2013

  52. [52]

    Trinquier, G

    J. Trinquier, G. Uguzzoni, A. Pagnani, F. Zamponi, and M. Weigt, Efficient generative modeling of protein sequences using simple autoregressive models, Nature communications12, 5800 (2021)

    2021

  53. [53]

    Morcos, A

    F. Morcos, A. Pagnani, B. Lunt, A. Bertolino, D. S. Marks, C. Sander, R. Zecchina, J. N. Onuchic, T. Hwa, and M. Weigt, Direct-coupling analysis of residue co- evolution captures native contacts across many protein families, Proceedings of the National Academy of Sci- ences108, E1293 (2011)

    2011

  54. [54]

    Cocco and R

    S. Cocco and R. Monasson, Adaptive cluster expansion for inferring boltzmann machines with noisy data, Phys- ical review letters106, 090601 (2011)

    2011

  55. [55]

    Barrat-Charlaix, A

    P. Barrat-Charlaix, A. P. Muntoni, K. Shimagaki, M. Weigt, and F. Zamponi, Sparse generative modeling via parameter reduction of boltzmann machines: appli- cation to protein-sequence families, Physical Review E 104, 024407 (2021)

    2021

  56. [56]

    Tsishyn, H

    M. Tsishyn, H. Talibart, M. Rooman, and F. Pucci, Structure-informed direct coupling analysis improves protein mutational landscape predictions, bioRxiv , 2026 (2026)

    2026

  57. [57]

    Haldane and R

    A. Haldane and R. M. Levy, Mi3-gpu: Mcmc-based inverse ising inference on gpus for protein covaria- tion analysis, Computer physics communications260, 107312 (2021)

    2021

  58. [58]

    Rosset, R

    L. Rosset, R. Netti, A. P. Muntoni, M. Weigt, and F. Zamponi, adabmDCA 2.0—a flexible but easy-to-use package for direct coupling analysis, inProtein Evolu- tion: Methods and Protocols, Methods in Molecular Bi- ology, Vol. 2979, edited by S. M. Khan and F. Pazos (Humana, New York, NY, 2026) pp. 83–104

    2026

  59. [59]

    W. P. Russ, M. Figliuzzi, C. Stocker, P. Barrat- Charlaix, M.Socolich, P.Kast, D.Hilvert, R.Monasson, S. Cocco, M. Weigt,et al., An evolution-based model for designing chorismate mutase enzymes, Science369, 440 (2020)

    2020

  60. [60]

    B. Fram, Y. Su, I. Truebridge, A. J. Riesselman, J. B. Ingraham, A. Passera, E. Napier, N. N. Thadani, S. Lim, K. Roberts,et al., Simultaneous enhancement of multiple functional properties using evolution-informed protein design, Nature communications15, 5141 (2024)

    2024

  61. [61]

    Fernandez-de Cossio-Diaz, P

    J. Fernandez-de Cossio-Diaz, P. Hardouin, F.-X. Ly- onnet du Moutier, A. Di Gioacchino, B. Marchand, Y. Ponty, B. Sargueil, R. Monasson, and S. Cocco, De- signing molecular rna switches with restricted boltz- mann machines, Nature Communications16, 11223 (2025)

    2025

  62. [62]

    Expanding functional protein sequence space using high entropy generative models

    R. Netti, E. Hinds, F. Calvanese, R. Ranganathan, M. Weigt, and F. Zamponi, Expanding functional pro- tein sequence space using high entropy generative mod- els, arXiv 2605.03578 (2026). 24

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  63. [63]

    Romero-Romero, S

    S. Romero-Romero, S. Lindner, and N. Ferruz, Explor- ing the protein sequence space with global generative models, Cold Spring Harbor Perspectives in Biology15, a041471 (2023)

    2023

  64. [64]

    Mardikoraem, Z

    M. Mardikoraem, Z. Wang, N. Pascual, and D. Woldring, Generative models for protein sequence modeling: recent advances and future directions, Briefings in Bioinformatics24, bbad358 (2023)

    2023

  65. [65]

    C. Hsu, H. Nisonoff, C. Fannjiang, and J. Listgarten, Learning protein fitness models from evolutionary and assay-labeled data, Nature biotechnology40, 1114 (2022)

    2022

  66. [66]

    Tubiana, S

    J. Tubiana, S. Cocco, and R. Monasson, Learning pro- tein constitutive motifs from sequence data, Elife8, e39397 (2019)

    2019

  67. [67]

    Shimagaki and M

    K. Shimagaki and M. Weigt, Selection of sequence mo- tifs and generative hopfield-potts models for protein families, Physical Review E100(2019)

    2019

  68. [68]

    Decelle, A

    A. Decelle, A. d. J. Navas Gómez, and B. Seoane, Infer- ring higher-order couplings with neural networks, Phys. Rev. Lett.135, 207301 (2025)

    2025

  69. [69]

    M. Huot, P. Rosenbaum, C. Planchais, H. Mouquet, R. Monasson, and S. Cocco, Generative model of sars-cov-2 variants under immune pressure unveils vi- ralescapepotential,bioRxiv10.1101/2025.05.12.653592 (2025)

    2025

  70. [70]

    M. Huot, D. Wang, E. Shakhnovich, R. Monasson, and S. Cocco, Constrained evolutionary funnels shape viral immune escape, Proceedings of the National Academy of Sciences123, e2536956123 (2026)

    2026

  71. [71]

    Rehan, E

    A. Rehan, E. Mauri, J. Fernandez-de Cossio-Diaz, P.-G. Brun, R. Monasson, M. Ribezzi-Crivellari, and S. Cocco, Design and experimental characterization of specificity-switching mutational paths of ww domains, bioRxiv 10.64898/2025.12.08.693000 (2025)

    work page doi:10.64898/2025.12.08.693000 2025

  72. [72]

    X. Ding, Z. Zou, and C. L. Brooks, Deciphering pro- tein evolution and fitness landscapes with latent space models, Nature Communications10, 10.1038/s41467- 019-13633-0 (2019)

    work page doi:10.1038/s41467- 2019

  73. [73]

    Hawkins-Hooker, F

    A. Hawkins-Hooker, F. Depardieu, S. Baur, G. Coua- iron, A. Chen, and D. Bikard, Generating functional protein variants with variational autoencoders, PLoS computational biology17, e1008736 (2021)

    2021

  74. [74]

    Ziegler, J

    C. Ziegler, J. Martin, C. Sinner, and F. Morcos, Latent generative landscapes as maps of functional diversity in protein sequence space, Nature Communications14, 10.1038/s41467-023-37958-z (2023)

    work page doi:10.1038/s41467-023-37958-z 2023

  75. [75]

    Shukla, J

    D. Shukla, J. Martin, F. Morcos, and D. A. Potoyan, Thermal adaptation of cytosolic malate dehydrogenase revealed by deep learning and coevolutionary anal- ysis, Journal of Chemical Theory and Computation 10.1021/acs.jctc.4c01774 (2025)

    work page doi:10.1021/acs.jctc.4c01774 2025

  76. [76]

    Rives, J

    A. Rives, J. Meier, T. Sercu, S. Goyal, Z. Lin, J. Liu, D. Guo, M. Ott, C. L. Zitnick, J. Ma,et al., Bio- logical structure and function emerge from scaling un- supervised learning to 250 million protein sequences, Proceedings of the National Academy of Sciences118, e2016239118 (2021)

    2021

  77. [77]

    Hesslow, N

    D. Hesslow, N. Zanichelli, P. Notin, I. Poli, and D.Marks,Rita: astudyonscalingupgenerativeprotein sequence models, arXiv 2205.05789 (2022)

    work page arXiv 2022

  78. [78]

    Sevgen, J

    E. Sevgen, J. Moller, A. Lange, J. Parker, S. Quigley, J. Mayer, P. Srivastava, S. Gayatri, D. Hosfield, C. Dilks,et al., Prot-vae: protein transformer varia- tional autoencoder for functional protein design, Pro- ceedings of the National Academy of Sciences122, e2408737122 (2025)

    2025

  79. [79]

    Notin, M

    P. Notin, M. Dias, J. Frazer, J. Marchena-Hurtado, A. N. Gomez, D. Marks, and Y. Gal, Tranception: pro- tein fitness prediction with autoregressive transformers and inference-time retrieval, inInternational Confer- ence on Machine Learning(PMLR, 2022) pp. 16990– 17017

    2022

  80. [80]

    A.Chandra, L.Tünnermann, T.Löfstedt,andR.Gratz, Transformer-based deep learning for predicting protein properties in the life sciences, Elife12, e82819 (2023)

    2023

Showing first 80 references.