pith. sign in

arxiv: 2604.23182 · v1 · submitted 2026-04-25 · 📡 eess.SY · cs.SY

An Exponentially stable Extended Kalman Filter with Estimate dependent Process noise Covariance for Chemical Reaction Networks

Suryasnata Dash , Abhishek Dey This is my paper

Pith reviewed 2026-05-08 07:44 UTC · model grok-4.3

classification 📡 eess.SY cs.SY
keywords Extended Kalman FilterChemical Langevin Equationchemical reaction networksstochastic stabilitybiomolecular systemsprocess noise covariancesampling periodgene expression model
0
0 comments X

The pith

An Extended Kalman Filter with Chemical Langevin Equation-based state-dependent process noise covariance yields exponentially mean-square bounded estimation errors for biomolecular systems below a sampling period bound.

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

The authors introduce an Extended Kalman Filter for partially known nonlinear stochastic biomolecular systems in which the process noise covariance is made dependent on the current state estimate using the Chemical Langevin Equation. This construction supplies the noise statistics directly from the underlying reaction rates instead of relying on constant heuristic values. They establish that the filter error is exponentially bounded in the mean-square sense whenever the sampling period does not exceed an explicitly derived upper limit. The method is tested on a nonlinear gene expression model, demonstrating stable performance without manual noise tuning. A reader would care because it replaces ad-hoc filter calibration with a kinetics-based design procedure that respects the stochastic character of chemical reaction networks.

Core claim

The central discovery is that equipping the Extended Kalman Filter with a process noise covariance matrix obtained by evaluating the Chemical Langevin Equation at the state estimate produces exponential mean-square stability of the estimation error for discrete-time chemical reaction networks, provided the sampling interval satisfies the derived upper bound.

What carries the argument

The state estimate-dependent process noise covariance matrix derived from the Chemical Langevin Equation, which is inserted into the covariance update equations to enable the subsequent stochastic stability analysis via mean-square boundedness proofs.

If this is right

  • The estimation error of the proposed filter converges exponentially in the mean-square sense under the sampling condition.
  • The framework applies to discrete-time biomolecular systems modeled by chemical reaction networks.
  • First-principles derivation of the noise covariance removes the need for heuristic tuning.
  • Simulation results on a gene expression model confirm the stability property in practice.

Where Pith is reading between the lines

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

  • This approach could support real-time state estimation in engineered genetic circuits without repeated retuning.
  • The sampling bound offers a criterion for selecting discretization steps in digital implementations of biomolecular controllers.
  • Similar state-dependent noise modeling might improve Kalman filters for other stochastic processes in systems biology.
  • Joint state-parameter estimation could become more robust by extending the same covariance construction.

Load-bearing premise

The biomolecular dynamics are faithfully represented by the Chemical Langevin Equation and the linearization performed by the Extended Kalman Filter remains sufficiently accurate over each sampling interval.

What would settle it

Numerical simulation of a chemical reaction network showing unbounded growth of the mean-square estimation error when the sampling period exceeds the derived upper bound, even with the state-dependent covariance in place.

Figures

Figures reproduced from arXiv: 2604.23182 by Abhishek Dey, Suryasnata Dash.

Figure 1
Figure 1. Figure 1: Filter error obtained for (a) RNA polymerase (b) mRNA transcript (c) ribosome and (d) protein states denotes ribosome and X indicates protein biomolecules. Both T and X are measurable with same noise distribution meaning C =  0 1 0 0 0 0 0 1 . With this example, we try to find a suitable maximum δ for which system would show stability. Here, Lf = 0.85, La = 0.8, v = 2.7657, CA = 100, view at source ↗
Figure 2
Figure 2. Figure 2: Temporal plot of (a) posterior error covariance matrix norm (b) mean square filter error norm system based and assumed values, we calculate coefficients of inequality in (15) to be 20274.482, 77497.805, 469.051 and −0.277 corresponding to descending order of power terms using the roots function in MATLAB. The positive root obtained is δmax = 0.542×10−3 i.e, δ < 0.00054. To verify filter error trajectory, w… view at source ↗
read the original abstract

Biomolecular systems are often modeled with partially known nonlinear stochastic dynamics, making state and parameter estimation a central challenge. While Kalman filtering techniques are widely used in this setting, their performance critically depends on the choice of the process noise covariance, which is typically assumed constant and heuristically tuned. Such assumptions are not justified for biomolecular systems, where intrinsic noise arises from underlying reaction kinetics. In this work, we propose an Extended Kalman Filter (EKF) with a state estimate-dependent process noise covariance based on Chemical Langevin Equation (CLE). Further, we analyze the stochastic stability of the proposed filter and derive conditions under which the estimation error remains exponentially bounded in the mean-square sense. In particular, we obtain an upper bound on the sampling period for discrete-time biomolecular systems that guarantees this property. The proposed framework is validated through simulations on a nonlinear gene expression model. This approach enables first principle-based modeling and filter design choices for synthetic biomolecular circuits, eliminating the need for heuristic tuning of the process noise covariance.

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

1 major / 1 minor

Summary. The manuscript proposes an Extended Kalman Filter (EKF) for biomolecular systems modeled by the Chemical Langevin Equation (CLE), with the process noise covariance made explicitly dependent on the current state estimate via the CLE diffusion term. It claims to derive conditions for exponential mean-square stability of the estimation error in the discrete-time setting and obtains an explicit upper bound on the sampling period that guarantees this boundedness. The framework is illustrated via simulations on a nonlinear gene-expression model.

Significance. If the central stability result holds, the work supplies a first-principles route to selecting the process-noise covariance in EKF designs for chemical reaction networks, removing the need for heuristic tuning. The sampling-period bound supplies concrete guidance for digital implementation of such estimators in synthetic-biology circuits. The approach aligns with the growing interest in stochastic control of biomolecular systems and could be extended to parameter estimation or feedback design.

major comments (1)
  1. [Stability analysis section (derivation of sampling-period bound)] The stochastic-stability analysis (presumably the section deriving the sampling-period bound) establishes exponential mean-square boundedness only for the linearized error system that incorporates the state-dependent CLE noise. The Taylor remainder terms arising from the nonlinear state-transition and measurement maps of the underlying CLE are not bounded explicitly in a manner independent of the sampling interval h. Consequently, it is not shown that the mean-square boundedness carries over to the true nonlinear filter when h approaches the derived upper limit; the remainders may grow with h and violate the stability claim even inside the stated bound.
minor comments (1)
  1. [Abstract] The abstract states that an upper bound on the sampling period is obtained but does not indicate its functional dependence on system parameters (e.g., reaction rates or noise intensity); a brief symbolic expression would improve readability.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the positive evaluation of the manuscript's significance and for the constructive comment on the stability analysis. We address the major concern point by point below.

read point-by-point responses

  1. Referee: The stochastic-stability analysis (presumably the section deriving the sampling-period bound) establishes exponential mean-square boundedness only for the linearized error system that incorporates the state-dependent CLE noise. The Taylor remainder terms arising from the nonlinear state-transition and measurement maps of the underlying CLE are not bounded explicitly in a manner independent of the sampling interval h. Consequently, it is not shown that the mean-square boundedness carries over to the true nonlinear filter when h approaches the derived upper limit; the remainders may grow with h and violate the stability claim even inside the stated bound.

    Authors: We appreciate this observation. The stability theorem is indeed stated for the full nonlinear discrete-time filter, but the proof proceeds by linearizing the error dynamics (using the CLE drift and diffusion) and then invoking a stochastic Lyapunov analysis on the resulting linear time-varying system with state-dependent noise. The Taylor remainders are implicitly controlled by the upper bound on h, since the discretization of the CLE itself is first-order accurate and the bound is chosen to keep the closed-loop error contraction stronger than the growth of discretization errors. However, we agree that an explicit, h-independent (or h-compatible) mean-square bound on the remainder terms—leveraging local Lipschitz constants of the reaction propensity functions—was not supplied in the current version. In the revised manuscript we will add a supporting lemma that derives such bounds (showing the remainder contribution is O(h) in mean-square norm uniformly on compact sets) and verifies that it remains a vanishing perturbation inside the stated sampling-period interval, thereby rigorously closing the gap between the linearized analysis and the nonlinear filter. revision: yes

Circularity Check

0 steps flagged

No circularity: derivation relies on external CLE model and standard EKF stability techniques

full rationale

The paper derives the process noise covariance directly from the Chemical Langevin Equation (an independent stochastic model of reaction kinetics) and applies standard mean-square stability analysis to obtain a sampling-period bound for the linearized EKF. No quoted step reduces a prediction or bound to a fitted parameter, self-definition, or self-citation chain; the central claims remain independent of the target result. The simulation on a gene-expression model serves as external validation rather than input to the derivation.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The proposal rests on the Chemical Langevin Equation as the source of state-dependent noise and on the validity of EKF linearization for the nonlinear stochastic dynamics; no free parameters or invented entities are mentioned in the abstract.

axioms (2)
  • domain assumption Biomolecular systems are accurately modeled by the Chemical Langevin Equation for their stochastic dynamics.
    Used as the basis for constructing the estimate-dependent process noise covariance.
  • domain assumption The nonlinear system permits a valid first-order linearization for the EKF.
    Required for the standard EKF update equations to apply.

pith-pipeline@v0.9.0 · 5481 in / 1348 out tokens · 38137 ms · 2026-05-08T07:44:49.133072+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

28 extracted references

  1. [1]

    Del Vecchio and R

    D. Del Vecchio and R. M. Murray, Biomolecular feedback systems . Prince- ton University Press, Princeton, NJ, 2015

    2015

  2. [2]

    The chemical Langevin equation,

    D. T. Gillespie, “The chemical Langevin equation,” The Journal of Chem- ical Physics, vol. 113, no. 1, pp. 297–306, 2000

    2000

  3. [3]

    Parameter estimation and model se lec- tion in computational biology,

    G. Lillacci and M. Khammash, “Parameter estimation and model se lec- tion in computational biology,” PLoS computational biology , vol. 6, no. 3, p. e1000696, 2010

    2010

  4. [4]

    State and parameter estimation of the h eat shock response system using Kalman and particle filters,

    X. Liu and M. Niranjan, “State and parameter estimation of the h eat shock response system using Kalman and particle filters,” Bioinformatics, vol. 28, no. 11, pp. 1501–1507, 2012

    2012

  5. [5]

    Uncertainty propagation f or deter- ministic models of biochemical networks using moment equations and t he extended Kalman filter,

    T. Kurdyaeva and A. Milias-Argeitis, “Uncertainty propagation f or deter- ministic models of biochemical networks using moment equations and t he extended Kalman filter,” Journal of the Royal Society Interface , vol. 18, no. 181, p. 20210331, 2021

    2021

  6. [6]

    Bayesian paramete r estima- tion for dynamical models in systems biology,

    N. J. Linden, B. Kramer, and P. Rangamani, “Bayesian paramete r estima- tion for dynamical models in systems biology,” PLoS computational biology, vol. 18, no. 10, p. e1010651, 2022

    2022

  7. [7]

    Robust Kalman filte r state estimation for gene regulatory networks,

    A. H. Abolmasoumi, M. Mohammadian, and L. Mili, “Robust Kalman filte r state estimation for gene regulatory networks,” IEEE/ACM Transactions on Computational Biology and Bioinformatics , vol. 20, no. 2, pp. 1395– 1405, 2022

    2022

  8. [8]

    H. J. Kushner, Stochastic Stability and Control . New York: Academic Press, Inc., 1967

    1967

  9. [9]

    Stability of linear and non-linear Kalman filters,

    T. Karvonen et al. , “Stability of linear and non-linear Kalman filters,” Master’s thesis, University of Helsinki , 2014

    2014

  10. [10]

    Conditions for asymptotic stability of th e dis- crete minimum-variance linear estimator,

    J. Deyst and C. Price, “Conditions for asymptotic stability of th e dis- crete minimum-variance linear estimator,” IEEE Transactions on Auto- matic Control , vol. 13, no. 6, pp. 702–705, 2003

    2003

  11. [11]

    Correction to

    J. Deyst, “Correction to” conditions for asymptotic stability o f the discrete minimum-variance linear estimator”,” IEEE Transactions on Automatic Control, vol. 18, no. 5, pp. 562–563, 2003

    2003

  12. [12]

    A. H. Jazwinski, Stochastic processes and filtering theory . New York: Dover Publications, 2007

    2007

  13. [13]

    Coping with singular transition ma trices in estimation and control stability theory,

    J. Moore and B. D. Anderson, “Coping with singular transition ma trices in estimation and control stability theory,” International Journal of Control , vol. 31, no. 3, pp. 571–586, 1980

    1980

  14. [14]

    Stochastic s tability of the discrete-time extended Kalman filter,

    K. Reif, S. Gunther, E. Yaz, and R. Unbehauen, “Stochastic s tability of the discrete-time extended Kalman filter,” IEEE Transactions on Automatic control, vol. 44, no. 4, pp. 714–728, 1999

    1999

  15. [15]

    On sta bility of a class of filters for nonlinear stochastic systems,

    T. Karvonen, S. Bonnabel, E. Moulines, and S. Sarkka, “On sta bility of a class of filters for nonlinear stochastic systems,” SIAM Journal on Control and Optimization , vol. 58, no. 4, pp. 2023–2049, 2020

    2023

  16. [16]

    Stochastic stability of Kalman-type nonlin ear fil- ters,

    S. Wei and J. C. Spall, “Stochastic stability of Kalman-type nonlin ear fil- ters,” in 2025 American Control Conference (ACC) , pp. 4461–4466, IEEE, 2025

    2025

  17. [17]

    How to not make the extended Kalman filter fail,

    R. Schneider and C. Georgakis, “How to not make the extended Kalman filter fail,” Industrial & Engineering Chemistry Research , vol. 52, no. 9, pp. 3354–3362, 2013

    2013

  18. [18]

    It’s a noisy business! genetic regu lation at the nanomolar scale,

    H. H. McAdams and A. Arkin, “It’s a noisy business! genetic regu lation at the nanomolar scale,” Trends in genetics , vol. 15, no. 2, pp. 65–69, 1999

    1999

  19. [19]

    Intrinsic noise in gene re gulatory networks,

    M. Thattai and A. Van Oudenaarden, “Intrinsic noise in gene re gulatory networks,” Proceedings of the National Academy of Sciences, vol. 98, no. 15, pp. 8614–8619, 2001

    2001

  20. [20]

    Extended Kalman filter for estimat ion of parameters in nonlinear state-space models of biochemical networ ks,

    X. Sun, L. Jin, and M. Xiong, “Extended Kalman filter for estimat ion of parameters in nonlinear state-space models of biochemical networ ks,” PloS one, vol. 3, no. 11, p. e3758, 2008

    2008

  21. [21]

    An extended Kalman filter with updated noise covariance for parameter estimation in chemical rea ction networks,

    S. Dash, A. S. S. Kumar, and A. Dey, “An extended Kalman filter with updated noise covariance for parameter estimation in chemical rea ction networks,” in 2025 11th International Conference on Control, Decision an d Information Technologies (CoDIT) , vol. 1, pp. 1577–1582, IEEE, 2025

    2025

  22. [22]

    A Kalman filter a pproach for biomolecular systems with noise covariance updating,

    A. Dey, K. Chakrabarti, K. K. Gola, and S. Sen, “A Kalman filter a pproach for biomolecular systems with noise covariance updating,” in 2019 Sixth Indian Control Conference (ICC) , pp. 262–267, IEEE, 2019

    2019

  23. [23]

    Simon, Optimal state estimation: Kalman, H infinity, and nonlinear approaches

    D. Simon, Optimal state estimation: Kalman, H infinity, and nonlinear approaches. New Jersey: John Wiley & Sons, 2006

    2006

  24. [24]

    Khalil, Nonlinear Systems

    H. Khalil, Nonlinear Systems . Pearson Education, New Jersey: Prentice Hall, 2002

    2002

  25. [25]

    Stability and positive supermartingales,

    R. Bucy, “Stability and positive supermartingales,” Journal of Differential Equations, vol. 1, no. 2, pp. 151–155, 1965

    1965

  26. [26]

    Observers for nonlinear stochastic s ystems,

    T.-J. Tarn and Y. Rasis, “Observers for nonlinear stochastic s ystems,” IEEE Transactions on Automatic Control , vol. 21, no. 4, pp. 441–448, 1976

    1976

  27. [27]

    The distribution of general quadratic f orms in norma,

    J. Dik and M. De Gunst, “The distribution of general quadratic f orms in norma,” Statistica Neerlandica, vol. 39, no. 1, pp. 14–26, 1985

    1985

  28. [28]

    A simple proof of Descartes’s rule of signs,

    X. Wang, “A simple proof of Descartes’s rule of signs,” The American Mathematical Monthly, vol. 111, no. 6, pp. 525–526, 2004

    2004