pith. sign in

arxiv: 2512.24427 · v2 · submitted 2025-12-30 · 🧬 q-bio.MN · nlin.AO· nlin.CD· physics.bio-ph· q-bio.QM

Epigenetic feedback reshapes dynamical landscapes in gene regulatory networks

Sascha H. Hauck , Sandip Saha , Narsis A. Kiani , Jesper N. Tegner This is my paper

Pith reviewed 2026-05-16 18:50 UTC · model grok-4.3

classification 🧬 q-bio.MN nlin.AOnlin.CDphysics.bio-phq-bio.QM
keywords gene regulatory networksepigenetic feedbackWaddington landscapedynamical mean field theorycell fate decisionsstochastic dynamicsdevelopmental reprogramming
0
0 comments X

The pith

Epigenetic feedback dynamically reshapes the Waddington landscape in gene regulatory networks.

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

The paper establishes a theoretical framework showing that slow epigenetic modifications in gene regulatory networks can reshape the effective potential landscape that guides cell states over time. It extends dynamical mean field theory by treating epigenetic marks as additional slow feedback variables, then uses the Hopfield network analogy to spin glasses to derive simpler stochastic equations that capture multi-timescale behavior. A sympathetic reader would care because the approach makes high-dimensional collective gene dynamics tractable enough to quantify how epigenetic changes create, remove, or tilt barriers between stable cell fates. This links molecular regulation directly to observable transitions in development and reprogramming without requiring full simulation of every interaction.

Core claim

The central claim is that epigenetic feedback regulation dynamically reshapes the Waddington landscape. The authors develop an extended Dynamical Mean Field Theory framework for gene regulatory networks that incorporates epigenetic modifications as slow, feedback-driven variables. Building on the analogy between Hopfield networks and spin glass systems, they derive effective stochastic equations that reduce high-dimensional dynamics to a tractable form across multiple timescales, enabling quantitative characterization of both stable and oscillatory regimes.

What carries the argument

Extended dynamical mean field theory (DMFT) that treats epigenetic modifications as slow feedback variables and reduces the high-dimensional GRN dynamics to effective stochastic equations via the Hopfield-spin glass analogy.

If this is right

  • Quantitative characterization of stable and oscillatory regimes in cell states becomes possible from the reduced equations.
  • Epigenetic feedback directly governs the creation or removal of barriers between cell fates in the effective landscape.
  • The framework unifies understanding of developmental dynamics and epigenetic reprogramming under one set of stochastic equations.
  • Analysis across fast gene-expression and slow epigenetic timescales can be performed without full high-dimensional simulation.

Where Pith is reading between the lines

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

  • The model suggests that interventions targeting epigenetic timescales could be used to steer cell fate trajectories in reprogramming protocols.
  • Extension to disease contexts could predict how altered epigenetic feedback destabilizes normal cell states in cancer or aging.
  • Predictions could be tested by overlaying measured epigenetic mark dynamics onto gene-expression time courses in differentiating cell populations.

Load-bearing premise

The high-dimensional dynamics of gene regulatory networks with epigenetic feedback can be reduced to tractable effective stochastic equations via the DMFT analogy to Hopfield networks and spin glasses without losing essential biological features.

What would settle it

Direct comparison of the model's predicted changes in the effective potential landscape against single-cell trajectories showing cell-state transitions when specific epigenetic modifiers are experimentally perturbed.

Figures

Figures reproduced from arXiv: 2512.24427 by Jesper N. Tegner, Narsis A. Kiani, Sandip Saha, Sascha H. Hauck.

Figure 1
Figure 1. Figure 1: Potentials for Regime 1 for different values of [PITH_FULL_IMAGE:figures/full_fig_p008_1.png] view at source ↗
Figure 4
Figure 4. Figure 4: Potentials for Regime 3 for different values of [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Dependence of Regime 3 on different values [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Potentials for Regime 4 for different values of [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Time evolution of θ for 1000 agents initially uniformly distributed in the range θ ∈ [−2.5, 2.5]. The density plots illustrate convergence to either one or two stable values depending on the parameter regime: (a) αβ < 1 and (b) αβ > 1, with c = 0. (19) with respect to τ , yielding,  − ∂ 2 ∂τ 2 − ∂ 2V ∂∆2  ∂ ∂τ ∆(τ ) = 0. Additionally, we obtain the eigenvalue to be equal to zero. However, it is still unc… view at source ↗
read the original abstract

Understanding how gene regulatory networks (GRNs) give rise to stable and dynamic cellular states remains a central challenge in theoretical biology, particularly when slow epigenetic feedback reshapes the underlying regulatory landscape. While experimental approaches such as single-cell transcriptomics reveal rich dynamical behaviour, a tractable theoretical framework that links gene expression, epigenetic control, and collective dynamics remains challenging. Here, we develop an extended Dynamical Mean Field Theory (DMFT) framework for GRNs that incorporates epigenetic modifications as slow, feedback-driven variables. Building on the analogy between Hopfield networks and spin glass systems, we derive effective stochastic equations that reduce high-dimensional dynamics to a tractable form across multiple timescales. This formulation enables quantitative characterization of both stable and oscillatory regimes and reveals how epigenetic feedback reshapes the effective potential landscape governing cell fate decisions. Our model shows how epigenetic feedback regulation dynamically reshapes the Waddington landscape. Our results and methodology provide a unified theoretical framework for understanding developmental dynamics and epigenetic reprogramming in complex biological systems.

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

2 major / 2 minor

Summary. The manuscript develops an extended Dynamical Mean Field Theory (DMFT) framework for gene regulatory networks that treats epigenetic modifications as slow, feedback-driven variables. Building on the Hopfield network and spin-glass analogy, it derives effective stochastic equations that reduce high-dimensional GRN dynamics to a tractable multi-timescale form, characterizes stable and oscillatory regimes, and claims that epigenetic feedback dynamically reshapes the effective potential (Waddington) landscape governing cell-fate decisions.

Significance. If the reduction steps are valid and preserve essential directed/asymmetric features of real GRNs, the work supplies a unified theoretical framework linking gene expression, epigenetic control, and collective dynamics. This could aid interpretation of single-cell transcriptomic data on developmental trajectories and reprogramming by providing explicit effective equations across timescales.

major comments (2)
  1. [DMFT derivation section] The central reduction to effective stochastic equations (described in the derivation following the Hopfield/spin-glass analogy) assumes symmetric couplings for closure of the mean-field equations, yet the manuscript supplies no explicit verification that directed, non-reciprocal regulatory edges remain compatible with the effective potential once slow epigenetic feedback is restored; this assumption is load-bearing for the claimed landscape reshaping.
  2. [Results on landscape reshaping] No explicit check (e.g., comparison of the reduced equations against direct simulation of an asymmetric GRN with slow epigenetic variables) is provided to confirm that the qualitative cell-fate phenomenology survives the approximation; without this, the reshaping result does not demonstrably follow from the high-dimensional model.
minor comments (2)
  1. [Abstract] The abstract describes the derivation and results but contains no explicit equations, key parameter definitions, or error bounds, which hinders immediate assessment of the reduction steps.
  2. [Methods] Notation for the epigenetic variables and their coupling to the fast GRN dynamics should be introduced with a clear table or diagram early in the methods to improve readability.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading and constructive comments on our manuscript. We address each major comment point by point below and have revised the manuscript to strengthen the presentation of the DMFT assumptions and to include additional validation.

read point-by-point responses

  1. Referee: The central reduction to effective stochastic equations (described in the derivation following the Hopfield/spin-glass analogy) assumes symmetric couplings for closure of the mean-field equations, yet the manuscript supplies no explicit verification that directed, non-reciprocal regulatory edges remain compatible with the effective potential once slow epigenetic feedback is restored; this assumption is load-bearing for the claimed landscape reshaping.

    Authors: We acknowledge that the derivation follows the standard symmetric-coupling closure of Hopfield/spin-glass DMFT. The manuscript does not contain an explicit verification that the effective potential remains well-defined for directed, non-reciprocal GRN edges once the slow epigenetic variables are restored. In the revised manuscript we have added a dedicated paragraph in the Methods section that (i) states the symmetry assumption explicitly, (ii) provides a perturbative argument showing that weak asymmetry is averaged by the slow epigenetic feedback to leading order, and (iii) discusses the regime in which strongly directed interactions would invalidate the potential description. We agree this clarification was necessary. revision: yes

  2. Referee: No explicit check (e.g., comparison of the reduced equations against direct simulation of an asymmetric GRN with slow epigenetic variables) is provided to confirm that the qualitative cell-fate phenomenology survives the approximation; without this, the reshaping result does not demonstrably follow from the high-dimensional model.

    Authors: We agree that a direct numerical comparison is the most convincing way to establish that the qualitative phenomenology survives the reduction. In the revised manuscript we have added a new supplementary figure that compares the effective DMFT trajectories against direct stochastic simulations of a small (N=20) asymmetric GRN with explicit slow epigenetic variables. The figure demonstrates that the locations of stable fixed points, the occurrence of oscillatory regimes, and the direction of landscape reshaping are preserved, while modest quantitative shifts in transition times are noted and discussed as a limitation of the mean-field closure. revision: yes

Circularity Check

0 steps flagged

No circularity: standard DMFT extension remains self-contained

full rationale

The paper extends the established Dynamical Mean Field Theory (DMFT) using the Hopfield-spin-glass analogy to incorporate slow epigenetic feedback as additional variables, deriving effective stochastic equations across timescales. No equations or steps are shown to reduce by construction to fitted inputs, self-definitions, or load-bearing self-citations; the landscape-reshaping result follows from the model assumptions and standard mean-field closure rather than tautological renaming or imported uniqueness theorems. The derivation is therefore independent of its own outputs.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The framework rests on the standard DMFT reduction for high-dimensional systems and the Hopfield-spin glass analogy, with epigenetic variables added as slow feedback; no explicit free parameters or invented entities are detailed in the abstract.

axioms (1)
  • domain assumption Gene regulatory networks can be modeled analogously to Hopfield networks and spin glass systems for deriving effective stochastic equations
    Invoked when reducing high-dimensional dynamics to tractable form across multiple timescales

pith-pipeline@v0.9.0 · 5500 in / 1135 out tokens · 34530 ms · 2026-05-16T18:50:32.809709+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

38 extracted references · 38 canonical work pages · 1 internal anchor

  1. [1]

    This analysis is further supported by the numerical simulations on the time-evolution ofθ i shown in Fig. 7. For timescales comparable or bigger than 1/ν, the in- fluence ofθbecomes noticeable and our potential ex- hibits an additional implicit time-dependency. For the stability analysis it makes sense to look at the connec- tion between the original stab...

    work page 1988

  2. [2]

    J. M. W. Slack.From egg to embryo: regional specifica- tion in early development. Cambridge University Press, 1991

    work page 1991

  3. [3]

    Furusawa and K

    C. Furusawa and K. Kaneko. A dynamical-systems view of stem cell biology.Science, 338(6104):215–217, 2012

    work page 2012

  4. [4]

    A Rand, A

    D. A Rand, A. Raju, M. S´ aez, F. Corson, and E. D Siggia. Geometry of gene regulatory dynamics.Proc. Nat. Aca. Sci. U.S.A., 118(38):e2109729118, 2021

    work page 2021

  5. [5]

    Raju and E

    A. Raju and E. D. Siggia. A geometrical model of cell fate specification in the mouse blastocyst.Development, 151(8):dev202467, 2024

    work page 2024

  6. [6]

    A. Raju, B. Xue, and S. Leibler. A theoretical perspec- tive on waddington’s genetic assimilation experiments. Proc. Nat. Aca. Sci. U.S.A., 120(51):e2309760120, 2023

    work page 2023

  7. [7]

    Fontaine, M

    M. Fontaine, M. J. Delas, et al. Dynamic landscape analysis of cell fate decisions: Predictive models of neural development from single-cell data.bioRxiv, pages 2025– 05, 2025

    work page 2025

  8. [8]

    D. J. Cislo, M. J. Del´ as, J. Briscoe, and E. D. Siggia. Re- constructing waddington’s landscape from data.bioRxiv, pages 2025–08, 2025

    work page 2025

  9. [9]

    Weinreb, A

    C. Weinreb, A. Rodriguez-Fraticelli, F. D. Camargo, and A. M. Klein. Lineage tracing on transcriptional land- scapes links state to fate during differentiation.Science, 367(6479):eaaw3381, 2020

    work page 2020

  10. [10]

    V. Garg, Y. Yang, et al. Single-cell analysis of bidirec- tional reprogramming between early embryonic states reveals mechanisms of differential lineage plasticities. bioRxiv, 2023. 13

    work page 2023

  11. [11]

    Matsushita and K

    Y. Matsushita and K. Kaneko. Homeorhesis in wadding- ton’s landscape by epigenetic feedback regulation.Phys. Rev. Res., 2:023083, 2020

    work page 2020

  12. [12]

    Matsushita, T

    Y. Matsushita, T. S Hatakeyama, and K. Kaneko. Dy- namical systems theory of cellular reprogramming.Phys- ical Review Research, 4(2):L022008, 2022

    work page 2022

  13. [13]

    K. Kaneko. Evolution of robustness to noise and muta- tion in gene expression dynamics.PLoS one, 2(5):e434, 2007

    work page 2007

  14. [14]

    Miyamoto, C

    T. Miyamoto, C. Furusawa, and K. Kaneko. Pluripo- tency, differentiation, and reprogramming: a gene ex- pression dynamics model with epigenetic feedback reg- ulation.PLoS computational biology, 11(8):e1004476, 2015

    work page 2015

  15. [15]

    C. H. Waddington.The Strategy of the Genes. Rout- ledge, 2014

    work page 2014

  16. [16]

    Yampolskaya, L

    M. Yampolskaya, L. Ikonomou, and P. Mehta. Find- ing signatures of low-dimensional geometric landscapes in high-dimensional cell fate transitions.bioRxiv, pages 2025–05, 2025

    work page 2025

  17. [17]

    Yampolskaya and P

    M. Yampolskaya and P. Mehta. Hopfield networks as models of emergent function in biology.arXiv preprint:2506.13076, 2025

    work page arXiv 2025

  18. [18]

    W. Wang, K. Ni, D. Poe, and J. Xing. Transiently in- creased coordination in gene regulation during cell phe- notypic transitions.PRX life, 2(4):043009, 2024

    work page 2024

  19. [19]

    Sompolinsky, A

    H. Sompolinsky, A. Crisanti, and H. J. Sommers. Chaos in random neural networks.Phys. Rev. Lett., 61:259– 262, Jul 1988

    work page 1988

  20. [20]

    Sompolinsky and A

    H. Sompolinsky and A. Zippelius. Relaxational dynam- ics of the edwards-anderson model and the mean-field theory of spin-glasses.Physical Review B, 25(11):6860, 1982

    work page 1982

  21. [21]

    Crisanti and H

    A. Crisanti and H. Sompolinsky. Dynamics of spin systems with randomly asymmetric bonds: Langevin dynamics and a spherical model.Physical Review A, 36(10):4922, 1987

    work page 1987

  22. [22]

    Rajan, LF Abbott, and H

    K. Rajan, LF Abbott, and H. Sompolinsky. Stimulus- dependent suppression of chaos in recurrent neural net- works.Physical review. E, Statistical, nonlinear, and soft matter physics, 82(1 Pt 1):011903, 2010

    work page 2010

  23. [23]

    Chaos-guided Input Structuring for Improved Learning in Recurrent Neural Networks

    P. Panda and K. Roy. Chaos-guided input structuring for improved learning in recurrent neural networks.arXiv preprint arXiv:1712.09206, 2017

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  24. [24]

    N. G. Van Kampen.Stochastic processes in physics and chemistry, volume 1. Elsevier, 1992

    work page 1992

  25. [25]

    S. H. Strogatz.Nonlinear dynamics and chaos: with ap- plications to physics, biology, chemistry, and engineer- ing. Chapman and Hall/CRC, 2024

    work page 2024

  26. [26]

    Engelken, F

    R. Engelken, F. Wolf, and L. F Abbott. Lyapunov spec- tra of chaotic recurrent neural networks.Physical Review Research, 5(4):043044, 2023

    work page 2023

  27. [27]

    On a formula for the product-moment coefficient of any order of a normal frequency distribu- tion in any number of variables.Biometrika, 12(1):134– 139, 1918

    Leonard Isserlis. On a formula for the product-moment coefficient of any order of a normal frequency distribu- tion in any number of variables.Biometrika, 12(1):134– 139, 1918

    work page 1918

  28. [28]

    K. R. Moon, D. Van Dijk, et al. Visualizing structure and transitions in high-dimensional biological data.Nat. Biotech., 37(12):1482–1492, 2019

    work page 2019

  29. [29]

    T¨ auber.Critical Dynamics: A Field Theory Approach to Equilibrium and Non-Equilibrium Scaling Behavior

    Uwe C. T¨ auber.Critical Dynamics: A Field Theory Approach to Equilibrium and Non-Equilibrium Scaling Behavior. Cambridge University Press, 2014

    work page 2014

  30. [30]

    P. C. Martin, E. D. Siggia, and H. A. Rose. Statistical dynamics of classical systems.Phys. Rev. A, 8:423–437, Jul 1973

    work page 1973

  31. [31]

    Pearson.Sturm–Liouville Theory: Past and Present

    David P. Pearson.Sturm–Liouville Theory: Past and Present. Birkh¨ auser Basel, 2005

    work page 2005

  32. [32]

    J. Xing. Reconstructing data-driven governing equa- tions for cell phenotypic transitions: integration of data science and systems biology.Physical Biology, 19(6):061001, 2022

    work page 2022

  33. [33]

    M. K. Jolly, M. Boareto, et al. Implications of the hybrid epithelial/mesenchymal phenotype in metastasis.Fron- tiers in Oncology, 5:155, 2015

    work page 2015

  34. [34]

    Matsushita and K

    Y. Matsushita and K. Kaneko. Generic optimization by fast chaotic exploration and slow feedback fixation. Physical Review Research, 5(2):023017, 2023

    work page 2023

  35. [35]

    Louis H. Y. Chen, Larry Goldstein, and Qi-Man Shao. Normal Approximation by Stein’s Method. Probability and Its Applications. Springer, Berlin, Heidelberg, 1 edi- tion, 2011. Appendix A: Treating Noise Expectation In this section, we present a detailed derivation of the noise expectation defined in Eq. (11). The objective is to evaluate the expectation va...

    work page 2011

  36. [36]

    (22), indeed satisfies Eq

    Proof thatV t(∆)acts as a potential for∆ In this appendix, we provide a detailed derivation showing that the effective potentialV t(∆), defined in Eq. (22), indeed satisfies Eq. (21). Our goal is to explicitly compute the derivative ofV t(∆) with respect to ∆ and show that it reproduces the right-hand side of Eq. (19). Starting from the definition, we hav...

    work page

  37. [37]

    sgn(∆) ˆα(τ) z1 − 1 ˆβ(τ) z3 # + sgn(∆)F h ˆα(τ)z1 + ˆβ(τ)z 3 + θt i +c i i F ′ ˆα(τ)z2 + ˆγ(τ)z3 + θt i +c i

    Derivation of the fluctuation potentialW The fluctuation potential is defined as W=− ∂2V t(∆) ∂∆2 . 17 We start by writing out the explicit form of the second derivative of the potential: ∂2V t(∆) ∂∆2 =−1− β2g2 2 lim N→∞ EN Z ∞ −∞ Z ∞ −∞ Z ∞ −∞ Dz3 Dz2 Dz1 × ( F ′ h ˆα(τ)z1 + ˆβ(τ)z 3 + θt i +c i i F ˆα(τ)z2 + ˆγ(τ)z3 + θt i +c i " sgn(∆) ˆα(τ) z1 − 1 ˆβ(...

    work page

  38. [38]

    Accordingly, we are interested in the long-time limit of the potential, V ±(∆)≡lim t→∞ V t(∆)

    Derivation ofV ±(θ) In this appendix, we derive the potential on timescales that are much longer than 1/ν, i.e., timescales sufficiently long forθ i to reach its extreme values. Accordingly, we are interested in the long-time limit of the potential, V ±(∆)≡lim t→∞ V t(∆). To simplify the analysis, we first note that the final values ofθ i can attain at mo...

    work page