Parameter Invariance Analysis of Moment Equations Using Dulmage-Mendelsohn Decomposition

Akito Igarashi, Yutaka Hori · 2025

Read Parameter Invariance Analysis of Moment Equations Using Dulmage-Mendelsohn Decomposition by Akito Igarashi, Yutaka Hori (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-2503-13336.txt.

Introduction

Akito Igarashi and Yutaka Hori hunt for steady traits in noisy chemical systems, in a 2025 paper. Living cells keep their behavior stable while the world shifts around them, a trait called homeostasis. For reactions ruled by chance, the math of average behavior often has too few fixed values to pin down every unknown. That looseness has blocked clean answers about which steady values ignore changes in the system's knobs.

The pair cut the tangle with a graph method that splits the system of laws into well-set blocks, stacked in a clear order. Walking that order shows which steady values cannot move when a knob turns, because their links to it are cut by structure alone. The method gives biologists a way to spot built-in stability in cell chemistry without solving the full messy system first.

Catalog record

CREATORAkito Igarashi, Yutaka Hori
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE36 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-13336

Browse the full library · About the Tower of Babel

Full text

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

Parameter Invariance Analysis of Moment Equations Using Dulmage-Mendelsohn Decomposition Akito Igarashi, Yutaka Hori Abstract Living organisms maintain stable functioning amid environmental fluctuations through homeostasis, a property that preserves a system’s behavior despite changes in environmental conditions. To elucidate homeostasis in stochastic biochemical reactions, theoretical tools for assessing population-level invariance under parameter perturbations are crucial. In this paper, we propose a systematic method for identifying the stationary moments that remain invariant under parameter perturbations by leveraging the structural properties of the stationary moment equations. A key step in this development is addressing the underdetermined nature of moment equations, which has traditionally made it difficult to characterize how stationary moments depend on system parameters. To overcome this, we utilize the Dulmage-Mendelsohn (DM) decomposition of the coefficient matrix to extract welldetermined subequations and reveal their hierarchical structure. Leveraging this structure, we identify stationary moments whose partial derivatives with respect to parameters are structurally zero, facilitating the exploration of fundamental constraints that govern homeostatic behavior in stochastic biochemical systems. I Introduction Homeostasis is the property by which organisms adapt to changes in external environments, enabling survival in highly uncertain conditions. Experimental studies have begun to reveal the underlying structural mechanisms of biomolecular reactions that enable homeostatic behaviors [ 1 , 2 , 3 ] . In this line of research, a recent study developed a synthetic biomolecular feedback system that is robust to disturbances [ 4 ] based on the theoretical analysis of stochastic biomolecular reaction networks. Mathematically, homeostasis can be described as the ability of the reaction network to maintain its state variables invariant to perturbations in parameters such as reaction rates, which may be affected by environmental factors such as temperature and osmolarity [ 4 , 1 , 5 ] . To date, many efforts have been made to analyze the invariance based on deterministic models, specifically the rate equations. These include approaches that utilize reaction stoichiometry to identify subnetworks attaining perfect adaptation [ 6 , 7 ] and those employing sensitivity analysis based on the Jacobian matrix [ 8 , 9 ] . However, since intracellular reactions are inherently stochastic due to the low copy number of molecules, developing methods based on stochastic models is desirable for a more accurate analysis that better aligns with experimental observations. The governing equation for stochastic chemical reactions is the chemical master equation (CME), an infinite-dimensional linear differential equation for a discrete-state Markov process that poses challenges for sensitivity and invariance analysis [ 10 , 11 ] . To address this, various numerical methods were proposed to evaluate the sensitivity of the density distributions of the copy numbers using the Fisher information matrix (FIM), employing Monte Carlo simulations or model approximations [ 12 , 13 ] . Meanwhile, moment-based approaches are amenable to analytic or algebraic characterization of stationary moments based on the moment equation [ 14 , 15 , 5 ] and enabled rigorous analysis of invariant moments for a certain class of biomolecular networks. A notable advancement in this direction is the mathematical characterization of the antithetic integral feedback (AIF) network [ 16 ] as a system that achieves maximal robust perfect adaptation (MaxRPA), a property in which the stationary moments are robust to all but two parameter perturbations [ 17 ] . Yet, a general framework is still lacking for systematically identifying stationary moments that remain invariant under arbitrary parameter perturbations. Addressing this gap is essential for advancing the theoretical understanding of homeostasis beyond known motifs. In this study, we propose a systematic method for analyzing the invariance of stationary moments with respect to arbitrary parameters by leveraging the structural information of the moment equations. We utilize the Dulmage-Mendelsohn (DM) decomposition [ 18 , 19 ] , a graph-based matrix operation that extracts welldetermined subequations from a system of linear equations, to identify determinable variables in the highly underdetermined stationary moment equations and the dependency structure among the variables. By linking this structure to unreachability analysis in directed graphs, we establish a computationally tractable algorithm to systematically identify stationary moments whose partial derivatives with respect to parameters are structurally zero. The proposed method contributes to a broader understanding and application of homeostasis by explicitly accounting for the stochastic nature of biochemical reactions. Notations: The symbol \mathbb{P}(\cdot) is a probability mass function, and \mathbb{E}[f(\cdot)] is the moment of the mass function defined by \sum_{\bm{x}}f(\bm{x})\mathbb{P}(\bm{x}) . The symbol \mathrm{supp}(\cdot) denotes the set of non-zero entries, i.e., \{(i,j)\mid(X)_{i,j}\neq 0\} for a matrix X and \{i\mid(\bm{x})_{i}\neq 0\} for a vector \bm{x} , respectively. The expression \bm{x}^{\bm{\alpha}} with \bm{x}=[x_{1},x_{2},\ldots,x_{n}]^{\top} and \bm{\alpha}=[\alpha_{1},\alpha_{2},\ldots,\alpha_{n}]^{\top} denotes a multivariate monomial of x_{1}^{\alpha_{1}}x_{2}^{\alpha_{2}}\dots x_{n}^{\alpha_{n}} . II Problem Formulation In this section, we introduce a model describing the stochastic dynamics of chemical reactions and formulate the parameter invariance analysis problem. II-A Chemical Master Equations We consider chemical reaction networks \Gamma consisting of n molecular species X_{i}\>(i=1,2,\dots,n) and m reaction species R_{j}\>(j=1,2,\dots,m) . In stochastic chemical reactions, the state of the reaction networks \Gamma is characterized by the copy number of molecular species X_{i} , and its evolution is modeled by the Markov process governed by stochastic changes in the copy numbers by reactions R_{j} . To formally define this stochastic process, let x_{i} denote the copy number of the molecular species X_{i} , and define the state of the reaction network \Gamma at time t by \bm{x}(t)=[x_{1},x_{2},\ldots,x_{n}]^{\top}\in\mathbb{N}_{0}^{n} . Reaction species R_{j} is characterized by a stoichiometry vector \bm{s}_{j}\in\mathbb{Z}^{n} , which describes the change in \bm{x} upon the occurrence of the reaction, and a propensity function w_{j}(\bm{x}) representing the probability of its occurrence per unit time. The reaction R_{j} is one of the elementary reactions, and the propensity function w_{j}(\bm{x}) is linear with respect to \theta_{j} and a polynomial in \bm{x} with a maximum degree of two [ 5 ] . The stochastic chemical reactions are then modeled as the evolution of the copy number distribution of the state \bm{x} at time t denoted by \mathbb{P}\left(\bm{x}|\bm{\theta},t,\bm{x}_{0}\right) , for a given initial state \bm{x}_{0} , i.e., \bm{x}_{0}:=\bm{x}(0) . The governing equation of the dynamics is called the Chemical Master Equation (CME) [ 10 , 11 ] . Omitting t and \bm{x}_{0} in \mathbb{P}\left(\bm{x}|\bm{\theta},t,\bm{x}_{0}\right) to avoid notational clutter, the CME can be expressed as \displaystyle\!\!\!\!\!\frac{d\mathbb{P}(\bm{x}|\bm{\theta})}{dt}\!=\! \displaystyle\sum_{j=1}^{m}\{w_{j}(\bm{x}\!-\!\bm{s}_{j})\mathbb{P}(\bm{x}\!-\!\bm{s}_{j}|\bm{\theta})\!-\!w_{j}(\bm{x})\mathbb{P}(\bm{x}|\bm{\theta})\}\!\!\!\! (1) for each \bm{x}\in\mathbb{N}_{0}^{n} , where \bm{\theta}:=[\theta_{1},\theta_{2},\ldots,\theta_{m}]^{\top}\in\mathbb{R}_{+}^{m} . It should be noticed that the CME is a linear but an infinite dimensional ordinary differential equation (ODE), for which analytic solutions are rarely available. II-B Problem statement In the design and analysis of stochastic reaction networks, one is often interested in understanding how the copy number distributions \mathbb{P}(\bm{x}|\bm{\theta}) depend on the rate parameters \bm{\theta} , which can be influenced by environmental changes such as temperature and osmolarity [ 5 ] . In this work, we examine these dependencies through the moments of the copy number distribution \mathbb{E}[\bm{x}^{\alpha}|\bm{\theta}] , which capture essential statistics such as the mean and the variance. More specifically, our focus is on the stationary moments, and we investigate whether certain stationary moments remain invariant under changes in the rate parameters \bm{\theta} , as such invariance is closely related to the homeostatic behavior of biomolecular systems. To formally define the problem, we introduce the following assumption, following literature [ 15 ] . Assumption 1 For any rate parameters \bm{\theta} and

Read the complete text