Noise-Driven Differentiation via Gene Frustration and Epigenetic Fixation
Davey Plugers, Kunihiko Kaneko · 2026
Read Noise-Driven Differentiation via Gene Frustration and Epigenetic Fixation by Davey Plugers, Kunihiko Kaneko (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-18185.txt.
Introduction
Plugers and Kaneko develop a theory for how noisy gene expression still yields reliable cell differentiation. Weakly stable intermediate states, called frustrated states, get nudged by random fluctuations and then locked in by slow epigenetic feedback. Removing the fast expression dynamics shows the slow epigenetic variable is driven by gene-expression noise, amplified by regulation. The theory predicts a logarithmic link between noise strength and differentiation onset time, plus input-dependent odds for each fate.
Catalog record
| CREATOR | Davey Plugers, Kunihiko Kaneko |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 30543 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.18185
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Noise-Driven Differentiation via Gene Frustration and Epigenetic Fixation Davey Plugers, Kunihiko Kaneko Abstract Gene expression in cells is stochastic, yet differentiation is robust. We propose a mechanism in which frustrated genes with weakly stable intermediate expression undergo noise-driven switching between basins of attraction, followed by irreversible fate fixation through slow epigenetic feedback. Regulatory interactions amplify effective noise and promote differentiation. We derive analytic expression for the logarithmic dependence of differentiation time on noise strength and input-dependent cell-fate selection, and demonstrate homeorhesis, the dynamical robustness of the epigenetic landscape. In biological systems, particularly within cells, stochasticity is unavoidable. As the number of molecules is not necessarily large and transcription often occurs in bursts, biochemical reactions and gene expression dynamics involve substantial stochasticity 7 ; 57 ; 51 ; 5 ; 55 ; 16 ; 40 ; 70 ; 27 ; 19 ; 24 ; 33 . Robustness to such noise is therefore expected to have evolved to enable cells to function properly. In fact, the noise tolerance of gene expression, responses to external signals, spatial patterning, and cell differentiation have been extensively investigated 12 ; 34 ; 43 ; 31 ; 38 ; 62 ; 65 . The positive roles of noise in cellular function have also been discussed 68 ; 15 , including noise- or discreteness-induced phase transitions 66 , stochastic resonance 41 , noise amplification 11 ; 54 , attractor selection by noise 54 ; 52 ; 30 ; 18 , boundary sharpening 75 , and cell-state exploration 67 ; 2 . Still, how cell differentiation can be both noise-driven and robust to noise remains elusive 48 ; 36 . In the classical Waddington landscape, differentiation over developmental time is described as a robust process termed homeorhesis 71 , in which cell states move along branching valleys. These branches correspond to differentiated cell types and their fractions, while the timing of branching is robust to intracellular stochasticity. Initially pluripotent cells can transition into various cell fates as intracellular gene-expression patterns and chromatin states change 49 ; 42 ; 32 . Such transitions from one to multiple stable states can be modeled through bifurcation mechanisms that represent the cell-fate decision process 26 ; 17 ; 39 ; 56 ; 59 ; 46 . However, gene expression levels within cells are noisy, and pluripotent cells must make reliable decisions near critical points for robust differentiation 47 ; 22 ; 46 ; 25 ; 2 . This requires suitable regulatory configurations and interactions to ensure that differentiation proceeds reproducibly 45 . Recently, we discovered that differentiation from a weakly stable gene-expression state, which we term a frustrated gene-expression state, can be achieved through noise together with a slow epigenetic process that increases stability as the system approaches bifurcated states 53 . In the present Letter, we analyze this cell-fate decision process based on frustrated gene-expression states, noise, and slow positive-feedback dynamics, and derive the conditions under which it occurs. We show that this differentiation process, although triggered by noise, is robust against noise, and we derive analytic expressions for the cell-state switching time and its dependence on the input. These results are supported by numerical simulations, which verify the theory and demonstrate the usefulness of the proposed mechanism for robust differentiation. Furthermore, we derive the Waddington epigenetic landscape over developmental time from the probability distribution of cellular states. We introduce a set of differential equations for the gene expression levels x_{i} and epigenetic factors \theta_{i} ( i=1,\dots,M ), following Ref. 44 (see also 60 ; 29 ). The change in the gene expression level x_{i} is determined by transcription F(y) , degradation -x_{i} , and a Gaussian noise term \sigma_{i}\eta_{i} that represents intracellular stochasticity in gene expression 55 ; 16 ; 40 ; 70 ; 27 ; 19 . Here, y=\sum_{j=1}^{M}J_{ij}x_{j}+\theta_{i}+c_{i}. The term J_{ij}x_{j} represents the regulatory interaction between genes: J_{ij}<0 means that gene j suppresses gene i , whereas J_{ij}>0 means that gene j activates gene i . Gene expression levels are attracted to on/off states. The fully expressed state x_{i}=1 is approached when \sum_{j}J_{ij}x_{j} exceeds the threshold -(\theta_{i}+c_{i}) ; otherwise, the unexpressed state x_{i}=-1 is approached 1 1 1 There are cases with limit cycles or chaotic attractors depending on J_{ij} . Differentiation from such states has already been studied in 53 and will not be discussed here. . Here, c_{i} is a constant activation offset, while \theta_{i} is a dynamic offset that starts from 0 and slowly approaches the value of x_{i} at a rate \nu\ll 1 through Eq. ( 1b ). \frac{dx_{i}}{dt}=F\left(\sum^{M}_{j}J_{ij}x_{j}+\theta_{i}+c_{i}\right)-x_{i}+\sigma_{i}\eta_{i} (1a) \frac{d\theta_{i}}{dt}=\nu(x_{i}-\theta_{i}) (1b) with Gaussian white noise satisfying \left<\eta_{i}(t),\eta_{j}(t^{\prime})\right>=\delta(t-t^{\prime})\delta_{ij} . We introduce the epigenetic factor \theta_{i} as in Ref. 39 , where it provides slow fixation of the system through a positive-feedback loop between gene expression and epigenetic modification 14 ; 61 . We initialize the system with \theta_{i}=0 at t=0 . Then, when a gene is expressed ( x_{i}>0 ), \theta_{i} gradually increases, which lowers the activation threshold -(\theta_{i}+c_{i}) and makes transcription of x_{i} more likely. The opposite occurs for unexpressed genes ( x_{i}<0 ). Transcription is determined by the activator function F(y)=\tanh(\beta y) , where \beta controls the steepness of the activation threshold. As the input moves away from the threshold ( y=0 ), the output of the activator function approaches a plateau, \tanh(\pm\infty)=\pm 1 , and the gene expression levels approach the corresponding values x_{i}=\pm 1 . In our previous work 53 , we found that the model consisting of Eqs. ( 1a ) and ( 1b ) can differentiate into bistable fixed points with stable gene expression levels and epigenetic factors, x_{i}=\theta_{i}=\pm 1 . The gene regulatory networks studied there were found to differentiate through frustrated genes, epigenetic fixation, and noise amplification 2 2 2 Another class of GRNs were also found which differentiated through successive stabilization of initial chaotic attractors to fixed points by the epigenetic fixation. . This was achieved by bringing the fixed point of the output genes close to x_{i}^{*}\approx 0 , which is the boundary between the basins of attraction for \theta_{i}=\pm 1 . Since expressed genes ( x_{i}>0 ) also provide positive feedback to \theta_{i} through d\theta_{i}/dt>0 , any fixed point eventually approaches x_{i}^{*}=+1 (or x_{i}^{*}=-1 if the gene remains unexpressed with x_{i}<0 ). Noise initially causes switching between the two basins of attraction, but the slow change in the epigenetic factors eventually induces irreversible cell-fate commitment, after which noise-driven switching between the basins is no longer possible. We refer to genes near such boundaries ( x_{i}\approx 0 ) as frustrated genes . They are termed frustrated because they remain at intermediate expression levels and have not yet committed to a final state. Such frustration can arise when signals that promote expression ( J_{is}>0 ) and signals that suppress expression ( J_{is}<0 ) balance each other. Frustrated genes can then be fixed to \pm 1 through noise and epigenetic fixation, or they can respond to inputs from signaling genes x_{s} . We begin by analyzing a single frustrated gene with c_{i}=0 and J_{ij}=0 for i\neq j . This gives a frustrated gene with x_{i}^{*}=\tanh(\beta J_{ii}x_{i}^{*})\Rightarrow x_{i}^{*}=0 . Here we require the system to be subcritical, \beta J_{ii}<1 , so that only a single stable fixed point