Macroscopic fluctuation-response theory and its use for gene regulatory networks

Timur Aslyamov, Krzysztof Ptaszyński, Massimiliano Esposito · 2025

Read Macroscopic fluctuation-response theory and its use for gene regulatory networks by Timur Aslyamov, Krzysztof Ptaszyński, Massimiliano Esposito (2025) 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-2510-17587.txt.

Introduction

Timur Aslyamov, with Krzysztof Ptaszyński and Massimiliano Esposito, links noise and response in living style networks in a 2025 arXiv paper. Many systems sit at a steady state that is not at rest, yet still jitter with random fluctuations. The team derives exact links between that jitter and how the system responds to a small push. Both sides of the link can be measured in experiments.

From those measures, one can rebuild the core parts of the system's linearized dynamics. The authors then apply the theory to gene regulatory networks, which are circuits where genes switch each other on and off. They split the noise spectrum into internal and external parts for any such network. Cross links between genes are included in the split.

Catalog record

CREATORTimur Aslyamov, Krzysztof Ptaszyński, Massimiliano Esposito
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE36 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2510-17587

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Macroscopic fluctuation-response theory and its use for gene regulatory networks Timur Aslyamov, Krzysztof Ptaszyński, Massimiliano Esposito Abstract Gaussian macroscopic fluctuation theory underpins the understanding of noise in a broad class of nonequilibrium systems. We derive exact fluctuation-response relations linking the power spectral density of stationary fluctuations to the linear response of stable nonequilibrium steady states. Both of these can be determined experimentally and used to reconstruct the kernel of the linearized dynamics and the diffusion matrix, and thus any features of the Gaussian theory. We apply our theory to gene regulatory networks with negative feedback, and derive an explicit internal-external noise decomposition of the power spectral density for any networks, including cross-correlations. Introduction— The behavior of many nonequilibrium systems can be modeled as small stochastic fluctuations around their deterministic dynamics. This type of behavior is most commonly described using linear Langevin equations, a framework that is ubiquitous across physics and applied mathematics elf2003fast ; paulsson2005models ; paulsson2004summing ; lestas2010fundamental ; yan2019kinetic ; bruggeman2009noise ; volpe2006torque ; alonso2007stochastic ; benayoun2010avalanches ; bressloff2010stochastic ; melbinger2012microtubule ; szavitz2014inherent ; dinis2012fluctuation ; han2021fluctuation ; aifer2024thermodynamic ; gilson2023entropy ; nicoletti2024tuning ; fyodorov2025nonorthogonal ; melanson2025thermodynamic . It can be derived in two complementary ways: either by linearizing the deterministic drift in a stochastic differential equation with additive Gaussian noise, an approach emphasized by van Kampen in his system-size expansion van1992stochastic ; kubo1966fluctuation ; gardiner2004handbook ; risken1989fokker ; stratonovich2012nonlinear , or by considering the macroscopic limit of an underlying Markov jump process and expanding around its most probable deterministic trajectory kampen1961power ; lax1960fluctuations ; keizer2012statistical ; kurtz1971limit . In both perspectives, the resulting dynamics of fluctuations is Gaussian and governed by an Ornstein–Uhlenbeck process, representing the universal description of small deviations near stable deterministic states. This regime may be viewed as a Gaussian macroscopic fluctuation theory, providing the lowest order approximation in the noise intensity of the general nonlinear theory of fluctuations around nonequilibrium steady states bertini2015macroscopic ; falasco2023macroscopic . Stationary fluctuations are characterized by two-point correlation functions, or their Fourier transform, the frequency-dependent Power Spectral Density (PSD)[ Eq. 5 ]. When integrated over all frequencies, the PSD reduces to the stationary covariance. While devoid of dynamical content, it quantifies the overall intensity of stationary fluctuations and satisfies the Lyapunov equation [ Eq. 7 ]. For systems obeying detailed balance (reciprocal dynamics), the fluctuation dissipation theorem kubo2012statistical provides an explicit relation for the stationary covariance in terms of the diffusion matrix and the static response of the system to perturbations [see Eq. 13 with \mathbb{Q}=1 ]. However, far from equilibrium, the lack of time-reversibility breaks this connection agarwal1972fluctuation ; marconi2008fluctuation ; seifert2010fluctuation ; prost2009generalized ; altaner2016fluctuation ; chun2023trade ; shiraishi2023introduction ; gao2024thermodynamic ; baiesi2009fluctuations ; baiesi2013update ; speck2016thermodynamic ; maes2020response ; tesser2024out ; klinger2025computing and a new nonequilibrium response theory is needed lucarini2016response ; santos2020response ; falasco2019negative ; mallory2020kinetic ; owen2020universal ; owen2023size ; gabriela2023topologically ; aslyamov2024nonequilibrium ; aslyamov2024general ; harunari2024mutual ; cengio2025mutual ; khodabandehlou2024affine ; floyd2024learning ; frezzato2024steady ; floyd2024limits ; gao2022thermodynamic ; zheng2025spatial ; auconi2025nonequilibrium ; floyd2025local ; ptaszynski2024dissipation ; kwon2024fluctuation . In this letter, we show that another quantity derived from the PSD, namely its zero-frequency component, can always be expressed in terms of the system responses and diffusion matrix, even in the absence of detailed balance. This relation is reminiscent of the Fluctuation-Response Relations (FRRs) recently discovered in the context of Markov jump processes aslyamov2024frr ; ptaszynski2024frr ; ptaszynski2025frr-mix ; aslyamov2025excess and of great practical value. As an application, we consider models of gene regulatory networks describing stationary fluctuations in mRNA and protein (with negative feedback) kaern2005stochasticity . Standard approaches study stationary correlation functions and are built on the Lyapunov equation paulsson2005models ; paulsson2004summing ; lestas2010fundamental ; yan2019kinetic ; bruggeman2009noise . More recent approaches use the PSD of the auto-correlation to capture dynamical features of stationary fluctuations simpson2003frequency ; warren2006exact ; gupta2022frequency ; song2019frequency . We provide an explicit expression for the full PSD matrix (including mRNA-protein cross-correlations) and show that it can be used to detect the presence of negative feedbacks. Most importantly, we show that our FRRs provide an unambiguous decomposition of the zero-frequency PSD in terms of intrinsic and extrinsic noise in arbitrary complex networks. This framework for decomposing noise draws inspiration from seminal experiments on gene expression variability elowitz2002stochastic ; pedraza2005noise . Figure 1: Linking the stationary fluctuations and responses far-from-equilibrium. Macroscopic dynamics— We consider a system described by the N -dimensional stochastic field \bm{x}(t) , which may correspond, e.g., to the particle position or a set of concentrations of chemical species. In the limit where the noise acting on the system becomes vanishingly small, the probability density of that field concentrates around the most likely value, \mathcal{X}(t) , which obeys the deterministic dynamical equation, \displaystyle d_{t}\bm{\mathcal{X}}(t) \displaystyle=\bm{f}(\bm{\mathcal{X}}(t))\,,\quad\bm{f}(\bm{x}^{*})=0\,, (1) where \bm{f}(\bm{\mathcal{X}}) is the rate vector and \bm{x}^{*} is a fixed point which we assume to be stable and unique. At steady state and in presence of a small but non-vanishing noise, the fluctuations of the stochastic field \bm{x}(t) in the vicinity of the fixed point are described by the linear Langevin equation \displaystyle d_{t}\bm{x}=\mathbb{K}(\bm{x}-\bm{x}^{*})+\sqrt{\varepsilon}\bm{\eta}\,. (2) Here, the first term describes the linear deterministic relaxation towards the fixed point, where \displaystyle\mathbb{K}\equiv\partial_{\bm{x}}\bm{f}(\bm{x})|_{\bm{x}=\bm{x}^{*}} (3) is the Jacobian of the deterministic dynamics with the elements K_{mk}=[\partial_{x_{k}}f_{m}(\bm{x})]_{\bm{x}=\bm{x}^{*}} for \bm{f}(\bm{x})=(\dots,f_{m}(\bm{x}),\dots)^{\intercal} and \bm{x}=(\dots,x_{k},\dots)^{\intercal} . The second term describes the effect of fluctuations as a Gaussian white noise vector \bm{\eta}(t)=(\ldots,\eta_{m}(t),\ldots)^{\intercal} with zero average \langle\bm{\eta}\rangle=0 and the autocorrelation function \displaystyle\langle\bm{\eta}(t)\bm{\eta}^{\intercal}(t^{\prime})\rangle=\delta(t-t^{\prime})\mathbb{D}\,, (4) where the average over noise realizations is denoted by \langle\dots\rangle and \mathbb{D} is the diffusion matrix that we assume to be positive semidefinite and constant. The prefactor \varepsilon is the noise amplitude, which corresponds, e.g., to the temperature for colloidal particles or the inverse volume in chemical systems. We note that \mathbb{K} and \mathbb{D} are often not known. We will see in what follows that they can be determined from the measurable linear response of arbitrary steady states (beyond equilibrium) and their stationary fluctuations, see Fig. 1 . Stationary fluctuations— Since Eq. 2 is linear, the dynamics of the average of \bm{x} is d_{t}\langle\bm{x}\rangle=\mathbb{K}(\langle\bm{x}\rangle-\bm{x}^{*}) , which is identical to the deterministic dynamics Eq. 1 linearized around the fixed point. Its fluctuations are captured by the two-point correlation functions \langle\delta\bm{x}(t+t^{\prime})\delta\bm{x}^{\intercal}(t^{\prime})\rangle , where \delta\bm{x}(t)=\bm{x}(t)-\langle\bm{x}(t)\rangle . The dynamical covariance is defined as \mathbb{C}(t)\equiv\varepsilon^{-1}\langle\delta\bm{x}(t)\delta\bm{x}^{\intercal}(t)\rangle . At steady state, the mean and covariance become time invariant, \langle\bm{x}(t)\rangle_{\text{ss}}=\bm{x}^{*} and \mathbb{C}(t)=\mathbb{C}_{\text{ss}} , and the correlation function becomes independent from t^{\prime} (which can thus be taken as t^{\prime}=0 ). The power spectrum density (PSD) of the steady state correlation function is \displaystyle\mathbb{Z}(\omega)\equiv\frac{1}{\varepsilon}\int_{-\infty}^{\infty}\langle\delta\bm{x}(t)\delta\bm{x}^{\intercal}(0)\rangle_{\text{ss}}e^{-i\omega t}dt\,, (5) where the scaling ensures that \mathbb{Z}(\omega) remains finite in the limit \varepsilon\rightarrow 0 . A known result keizer2012statistical , that we re-derive for completeness in Appendix A , is that \displaystyle\mathbb{Z}(\omega)=(\mathbb{K}-i\omega\mathbb{1})^{-1}\mathbb{D}[(\mathbb{K}-i\omega\mathbb{1})^{-1}]^{\dagger}\,, (6) where A^{\dagger}\equiv(\overline{A})^{\intercal} denotes the adjoint operation. Two important and complementary quantities can be derived from it. First, the zero frequency PSD , \mathbb{Z}(0) , which by time integrating the steady state correlation function, captures the persistence of fluctuations berg1977physics . Second, the stationary covariance which is obtained from it as \mathbb{C}_{\text{ss}}=(2\pi)^{-1}\int_{-\infty}^{\infty}d\omega\mathbb{Z}(\omega) and measures the magnitude and the directionality of fluctuations around the steady state. It can be calculated as the steady-state solution of the Lyapunov equation \displaystyle\mathbb{K}\mathbb{C}_{\text{ss}}+\mathbb{C}_{\text{ss}}\mathbb{K}^{\intercal}=-\mathbb{D}\,. (7) Efficient

Read the complete text