Information geometry of chemical reaction networks: Cramer-Rao bound and absolute sensitivity revisited
Dimitri Loutchko, Yuki Sughiyama, Tetsuya J. Kobayashi · 2025
Read Information geometry of chemical reaction networks: Cramer-Rao bound and absolute sensitivity revisited by Dimitri Loutchko, Yuki Sughiyama, Tetsuya J. Kobayashi (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-2503-19384.txt.
Introduction
Dimitri Loutchko, Yuki Sughiyama, and Tetsuya Kobayashi rebuild the information geometry of chemical reaction networks on Riemannian ground. Their 2025 paper starts from Legendre duality. That duality links concentration space and potential space, with stoichiometry as a constraint, and its maps act as isometries. The same holds for the natural parametrizations of the equilibrium subspace. They compare two metric tensors, and that step yields a multivariate Cramer-Rao bound.
A newer notion, absolute sensitivity, comes next. It is recast as a projection onto the tangent bundle of the equilibrium manifold. The authors give a linear algebra account of it. They also give first corrections for ideal solutions. A test case is the glyoxylate bypass system that governs IDHKP and IDH. In that case, hypersensitivity and negative self-regulation emerge from thermodynamic tuning alone. No exotic kinetics are needed. The target is the crowded chemistry inside cells.
Catalog record
| CREATOR | Dimitri Loutchko, Yuki Sughiyama, Tetsuya J. Kobayashi |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 106 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-19384
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Information geometry of chemical reaction networks: Cramer-Rao bound and absolute sensitivity revisited Dimitri Loutchko, Yuki Sughiyama, Tetsuya J. Kobayashi Abstract Information geometry is based on classical Legendre duality but allows to incorporate additional structure such as algebraic constraints and Bregman divergence functions. It is naturally suited, and has been successfully used, to describe the thermodynamics of chemical reaction networks (CRNs) based on the Legendre duality between concentration and potential spaces, where algebraic constraints are enforced by the stoichiometry. In this article, the Riemannian geometrical aspects of the theory are explored. It is shown that duality between concentration and potential spaces and the natural parametrizations of equilibrium subspace are isometries, which leads to a multivariate Cramer-Rao bound through the comparison of two Riemannian metric tensors. In the subsequent part, the theory is applied to the recently introduced concept of absolute sensitivity. Using the Riemannian geometric tools, it is proven that the absolute sensitivity is a projection operator onto the tangent bundle of the equilibrium manifold. A linear algebraic characterization and explicit results on first order corrections to the thermodynamics of ideal solutions are provided. Finally, the theory is applied to the IDHKP-IDH glyoxylate bypass regulation system. The novelty of the theory is that it is applicable to CRNs with non-ideal thermodynamical behavior, which are prevalent in highly crowded cellular environments due to various interactions between the chemicals. Indeed, the analyzed example shows remarkable behavior ranging from hypersensitivity to negative-self regulations. These are effects which usually require strongly nonlinear reaction kinetics. However, here, they are obtained by tuning thermodynamical interactions providing a complementary, and physically well-founded, viewpoint on such phenomena. 1 Introduction The theory of chemical reaction networks (CRNs) plays a central role in the understanding of biomolecular processes and mechanisms such as circadian clocks [ 21 , 23 ] , proofreading kinetics [ 26 , 9 ] , absolute concentration robustness [ 10 , 4 , 8 , 30 , 43 , 31 , 40 ] , perfect adaptation [ 24 , 32 ] , hypersensitivity [ 1 , 54 , 44 ] , among many others [ 41 , 3 ] . The success of CRN theory is rooted in its well-founded and constantly evolving mathematical basis. Initiated by Horn, Jackson, and Feinberg [ 28 , 17 , 27 , 19 ] , the contemporary mathematics of CRN is firmly rooted in the theory of differential equations [ 18 , 42 ] , graph theory [ 12 , 15 ] and algebraic geometry [ 11 , 29 , 13 ] but also spreads across various other disciplines such as homological algebra [ 25 ] and, most recently, information geometry. The information geometrical approach makes use of the analogy between the vector of chemical concentrations and probability distributions on finite spaces and allows to transfer techniques from statistics and information theory to study CRN. This viewpoint has allowed to establish a geometrical understanding of the thermodynamics of CRN [ 52 , 34 , 33 ] and provided new insights into thermodynamic uncertainty relations and speed limits [ 37 ] , growing systems [ 51 ] , and finite time driving [ 36 ] . In this article, this line of work is extended by exploring the Riemannian geometrical aspects of the information geometry of CRNs. All spaces are equipped with strictly convex potential functions inducing Legendre duality, and the natural Riemannian metrics are provided by the Hessians of the respective convex functions. It is then shown that Legendre transformations are isometries with respect to this structure. In fact, the Riemannian metric tensor can be thought of as encoding the information on the derivative of the Legendre transformation at each point. This compatibility of derivatives and metric tensors is used throughout the text. With regard to CRN theory, the relationships between the concentration and potential spaces, the respective equilibrium subspaces and their natural parameter spaces are established and condensed into commutative diagrams. It is then shown that these diagrams are diagrams of isometries which leads to formulation of a multivariate Cramer-Rao bound for CRNs as the comparison of two Riemannian metric tensors. In the second part, the concept of absolute sensitivity , which was introduced in [ 38 ] , is analyzed through the lens of the Riemannian geometric results. The classical sensitivity matrix encodes how the steady state concentrations of a CRN change upon infinitesimal perturbations of the linear conserved quantities. It contains valuable information on the robustness of the CRN but the numerical values of the matrix elements depend on the choice of a basis of \mathrm{Ker}[S^{*}] (where S is the stoichiometric matrix of the respective CRN) and are therefore difficult to interpret. In [ 38 ] , a construction was provided which retains the function of the sensitivity matrix but rectifies the dependence on the choice of a basis. The essential idea is to trace how an infinitesimal perturbation \delta x^{i} in the concentration of a chemical X_{i} of a state x changes the steady state and thus propagates to changes of the concentrations A^{j}_{i}(x)\delta x^{i} of all other chemicals X_{j} in the adjusted steady state. The quantity A^{j}_{i}(x) is called the absolute sensitivity of X_{j} with respect to X_{i} . Such ideas have also been explored from the viewpoint of computational algebra in [ 20 ] . In [ 38 ] , a linear algebraic characterization of absolute sensitivity was provided for quasi-thermostatic CRNs via a linear algebraic Cramer-Rao bound. This article puts this characterization into a proper information geometric context by promoting the local linear algebra to global tensor algebra and the matrix of absolute sensitivities to a projection operator, respectively. Most importantly, it becomes possible to study the geometry induced by potential functions of non-ideal systems 1 1 1 Quasi-thermostatic CRNs are precisely the CRNs which have the geometry of equilibrium CRNs with the thermodynamics of ideal solutions. . In doing so, the scope of applicability is greatly enhanced as the effects of crowded environments with, e.g., volume exclusion effects as well as attractive and repulsive interactions between the chemicals can be analyzed. This is important for biochemical applications, because cellular environments are far from ideal, and yet the CRNs of in vivo systems are usually described by mass action kinetics and ideal potential functions. Using the example of the core module of the IDHKP-IDH glyoxylate bypass regulation system, it is demonstrated how kinetically unusual but biologically important effects can be the result of thermodynamical interactions in a crowded environment. The article is structured as follows: In the remainder of this introductory section, the mathematical notation is introduced and an outline of the main constructions and results is given. In Sections 2 and 3 , the basic notions of CRN theory and Legendre duality are introduced, and examples of equilibrium manifolds are given. Section 4 is the main mathematical section of this article as it contains the analysis of Riemannian geometrical aspects in 4.1 and the proof of the Cramer-Rao bound in 4.2 . This is used to derive the main results on absolute sensitivity as a projection operator in Section 5 , followed by an analysis of first order perturbations of the geometry induced by ideal potential functions in Section 6 . An example is