pith. machine review for the scientific record. sign in

arxiv: 2605.07614 · v1 · submitted 2026-05-08 · 🧮 math.DS · q-bio.MN

Recognition: 2 theorem links

· Lean Theorem

Predictive-Switching Control of Stochastic Gene Regulatory Networks: A Contractive PIDE Framework

Christian Fern\'andez, G\'abor Szederk\'enyi, Irene Otero-Muras, Manuel P\'ajaro

Pith reviewed 2026-05-11 02:52 UTC · model grok-4.3

classification 🧮 math.DS q-bio.MN
keywords predictive switching controlgene regulatory networkspartial integro-differential equationsL1-contractivitystochastic stabilityswitching systemsneural network control
0
0 comments X

The pith

Predictive switching control renders the PIDE dynamics of gene regulatory networks L1-contractive, so the probability density forgets its initial condition.

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

The paper develops a predictive switching controller that selects inputs from a finite candidate set to steer the shape of the probability density function in stochastic gene networks governed by a PIDE model. Inputs are chosen at each step to minimize a given cost, and a hybrid neural-network version is introduced for computational scaling in higher dimensions. The central result is a contraction argument showing that the closed-loop PIDE operator is contractive in the L1 norm, which implies that solutions from different initial densities converge to each other. When leakage terms are strictly positive the contraction becomes exponential, supplying explicit stability certificates. The claims are supported by numerical simulations on networks of increasing dimension.

Core claim

Under the proposed predictive-switching law the closed-loop PIDE is L1-contractive; consequently the probability density of the gene-network state evolves independently of the initial distribution. With strictly positive leakage the contraction is exponential, furnishing rigorous asymptotic stability for the controlled stochastic process.

What carries the argument

L1-contractivity of the closed-loop PIDE operator under predictive switching, which forces the probability density to become independent of the initial measure.

If this is right

  • The probability density function can be driven toward prescribed shapes by discrete switching decisions.
  • Formal stability holds without requiring knowledge of the exact initial distribution.
  • Exponential convergence rates are obtained whenever leakage terms remain strictly positive.
  • Higher-dimensional instances remain tractable through the hybrid neural-network policy approximation.
  • The same contractivity analysis applies to the three representative network examples simulated in the paper.

Where Pith is reading between the lines

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

  • The same contraction argument could be tested on other integro-differential models of stochastic biological processes.
  • Model mismatch between the PIDE and actual gene expression data would be the primary obstacle to transferring the guarantees to laboratory experiments.
  • If leakage can be tuned experimentally, the exponential rate result supplies a quantitative design target for synthetic circuits.

Load-bearing premise

The gene regulatory network dynamics are accurately captured by the chosen PIDE model and that a finite candidate set of control inputs contains near-optimal choices.

What would settle it

Two trajectories of the closed-loop PIDE started from distinct initial probability densities whose L1 distance fails to decrease monotonically under the switching policy.

Figures

Figures reproduced from arXiv: 2605.07614 by Christian Fern\'andez, G\'abor Szederk\'enyi, Irene Otero-Muras, Manuel P\'ajaro.

Figure 1
Figure 1. Figure 1: Schematic of the gene regulatory network. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Relationship between the fine numerical grid [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: (a) Input signal. Top: final interval. Bottom: full [PITH_FULL_IMAGE:figures/full_fig_p012_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Contractivity analysis for Case Study I. (a) Pair 1 [PITH_FULL_IMAGE:figures/full_fig_p013_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: (a) Input signals. Top: final interval. Bottom: [PITH_FULL_IMAGE:figures/full_fig_p013_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Contractivity analysis for Case Study II. (a) Pair 1 [PITH_FULL_IMAGE:figures/full_fig_p014_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Stationary probability distribution of the uncon [PITH_FULL_IMAGE:figures/full_fig_p014_7.png] view at source ↗
Figure 9
Figure 9. Figure 9: Contractivity analysis for Case Study III. (a) Pair 1 [PITH_FULL_IMAGE:figures/full_fig_p015_9.png] view at source ↗
read the original abstract

This paper develops a predictive switching control algorithm for stochastic gene regulatory networks described by a Partial Integro-Differential Equation (PIDE) model, which enables direct shape control of the probability density function. Control inputs are selected from a finite candidate set to minimize a prescribed cost functional. A hybrid framework is proposed for scalability in higher-dimensional systems, using neural networks to approximate the control policy. A central theoretical contribution is a contraction-based analysis of the closed-loop PIDE dynamics. The paper establishes $L^ 1$-contractivity under the proposed control scheme, yielding formal stability guarantees and showing that the evolution of the probability density becomes progressively independent of the initial condition. Moreover, under strictly positive leakage terms, exponential convergence is obtained. The effectiveness and flexibility of the approach, together with the theoretical contractivity results, are illustrated through numerical simulations on three representative examples of increasing dimensionality.

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

2 major / 2 minor

Summary. The paper develops a predictive-switching control algorithm for stochastic gene regulatory networks modeled by partial integro-differential equations (PIDEs). Control inputs are selected from a finite candidate set to minimize a cost functional, with a hybrid neural-network approximation proposed for scalability in higher dimensions. The central theoretical contribution is a contraction-based analysis establishing L1-contractivity of the closed-loop PIDE dynamics, which yields formal stability guarantees, progressive independence of the probability density evolution from the initial condition, and (under strictly positive leakage terms) exponential convergence. Effectiveness is illustrated via numerical simulations on three representative examples of increasing dimensionality.

Significance. If the L1-contractivity result holds with the stated qualifiers, the work supplies a rigorous, contraction-based framework for direct shape control of probability densities in stochastic GRNs, offering formal initial-condition-independent stability guarantees that are uncommon in this domain. The hybrid NN extension addresses practical scalability, and the finite-candidate-set formulation keeps the approach computationally tractable. These elements, together with the explicit leakage-term condition for exponential rates, constitute a substantive advance for control-theoretic approaches to systems biology.

major comments (2)
  1. [Abstract] Abstract: The central claim of L1-contractivity and exponential convergence is asserted without any derivation outline, error bounds, or verification steps. Because this contraction result is the load-bearing theoretical contribution, the absence of these details prevents assessment of whether the proof is complete and whether the stated qualifiers (finite candidate set, positive leakage) are handled rigorously.
  2. [Numerical simulations] Numerical simulations section: The three examples are described as illustrating effectiveness and flexibility, yet no quantitative metrics (e.g., L1-error decay rates, cost values, or comparisons against open-loop or alternative controllers) are supplied. Without such evidence the practical support for the theoretical guarantees cannot be evaluated.
minor comments (2)
  1. [Abstract] Abstract: The notation “$L^ 1$” contains an extraneous space; standard mathematical typesetting uses $L^1$.
  2. [Abstract] Abstract: The description of the hybrid framework would benefit from a brief statement of how the neural-network policy approximation interfaces with the finite candidate set and the PIDE solver.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the thoughtful and constructive comments on our manuscript. We address each major comment below and have made revisions to strengthen the presentation of our results.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The central claim of L1-contractivity and exponential convergence is asserted without any derivation outline, error bounds, or verification steps. Because this contraction result is the load-bearing theoretical contribution, the absence of these details prevents assessment of whether the proof is complete and whether the stated qualifiers (finite candidate set, positive leakage) are handled rigorously.

    Authors: The abstract serves as a concise summary of the paper's contributions and results, consistent with standard academic practice where detailed derivations are reserved for the main text. The full proof of the L1-contractivity, including rigorous handling of the finite candidate set for control inputs and the positive leakage terms for exponential convergence, is presented in Section 3 (Theorem 3.1 and Corollary 3.2). No additional error bounds are introduced beyond the contractivity mapping, as the result follows directly from the PIDE analysis under the model assumptions. To facilitate assessment, we have revised the abstract to include a brief outline of the contraction argument and explicit reference to the theorem. revision: partial

  2. Referee: [Numerical simulations] Numerical simulations section: The three examples are described as illustrating effectiveness and flexibility, yet no quantitative metrics (e.g., L1-error decay rates, cost values, or comparisons against open-loop or alternative controllers) are supplied. Without such evidence the practical support for the theoretical guarantees cannot be evaluated.

    Authors: We agree that quantitative metrics would strengthen the numerical validation. In the revised manuscript, we have augmented the Numerical Simulations section with tables reporting L1-error norms at successive time steps for each example, the achieved cost functional values under the predictive-switching policy, and direct comparisons to open-loop evolution as well as a standard proportional feedback controller. These metrics confirm the exponential decay rates predicted by the theory when leakage terms are positive and demonstrate superior performance of the proposed method. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected in derivation chain

full rationale

The paper presents a predictive-switching control for PIDE-modeled gene networks, with the central contribution being a contraction-based analysis establishing L1-contractivity of the closed-loop dynamics. This yields stability guarantees and initial-condition independence (with exponential convergence under positive leakage). No load-bearing step reduces by construction to fitted inputs, self-definitions, or self-citation chains; the contractivity result is framed as an independent theoretical analysis with explicit assumptions and qualifiers. The derivation chain remains self-contained against the stated model and control scheme.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The framework rests on standard modeling assumptions for stochastic GRNs and control theory; no free parameters or invented entities are explicitly introduced in the abstract.

axioms (2)
  • domain assumption Gene regulatory network dynamics can be represented by a PIDE model
    Central modeling choice stated in the abstract
  • domain assumption Control inputs are restricted to a finite candidate set
    Required for the predictive switching algorithm and scalability claim

pith-pipeline@v0.9.0 · 5468 in / 1243 out tokens · 37604 ms · 2026-05-11T02:52:11.530884+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

Reference graph

Works this paper leans on

64 extracted references · 64 canonical work pages

  1. [1]

    Cañizo and José A

    José A. Cañizo and José A. Carrillo and Manuel Pájaro , issue =. Journal of Mathematical Biology , month =

    work page

  2. [2]

    Physics of Life Reviews , volume =

    Paulsson, Johan , title =. Physics of Life Reviews , volume =

    work page

  3. [3]

    Stochastic modeling and numerical simulation of gene regulatory networks with protein bursting , journal =

    P. Stochastic modeling and numerical simulation of gene regulatory networks with protein bursting , journal =

    work page

  4. [4]

    Bioinformatics , volume =

    P. Bioinformatics , volume =

    work page

  5. [5]

    A Kinetic Finite Volume Discretization of the Multidimensional

    V. A Kinetic Finite Volume Discretization of the Multidimensional. Bulletin of Mathematical Biology , volume =

    work page

  6. [6]

    Transient hysteresis and inherent stochasticity in gene regulatory networks , journal =

    P. Transient hysteresis and inherent stochasticity in gene regulatory networks , journal =

    work page

  7. [7]

    1983 , series =

    Pazy, Amnon , title =. 1983 , series =

    work page 1983

  8. [8]

    On the perturbation theory for strongly continuous semigroups , journal =

    J. On the perturbation theory for strongly continuous semigroups , journal =. 1977 , volume =

    work page 1977

  9. [9]

    2000 , series =

    Engel, Klaus-Jochen and Nagel, Rainer , title =. 2000 , series =

    work page 2000

  10. [10]

    and Tweedie, Richard L

    Meyn, Sean P. and Tweedie, Richard L. , title =

    work page

  11. [11]

    , title =

    Hairer, Martin and Mattingly, Jonathan C. , title =. Seminar on Stochastic Analysis, Random Fields and Applications VI , pages =

    work page

  12. [12]

    Davis, M. H. A. , title =. 1993 , series =

    work page 1993

  13. [13]

    Qualitative properties of certain piecewise deterministic

    Bena. Qualitative properties of certain piecewise deterministic. Annales de l’Institut Henri Poincar

    work page

  14. [14]

    , title =

    Harris, Theodore E. , title =. Proceedings of the Third Berkeley Symposium on Mathematical Statistics and Probability , volume =. 1956 , publisher =

    work page 1956

  15. [15]

    Tohoku Mathematical Journal , year =

    Isao Miyadera , title =. Tohoku Mathematical Journal , year =

    work page

  16. [16]

    Deep Learning , author=

    work page

  17. [17]

    Data-Driven Science and Engineering: Machine Learning, Dynamical Systems, and Control , author=

    work page

  18. [18]

    IEEE Transactions on Computational Biology and Bioinformatics , year =

    Faquir, Hamza and P\'ajaro, Manuel and Otero-Muras, Irene , title =. IEEE Transactions on Computational Biology and Bioinformatics , year =

    work page

  19. [19]

    PLoS Computational Biology , volume =

    Modi, Shakti and Dey, Supravat and Singh, Abhyudai , title =. PLoS Computational Biology , volume =

    work page

  20. [20]

    Nature Communications , volume =

    Lugagne, Jean-Baptiste and Sosa Carrillo, Sebasti\'an and Kirch, Melanie and K\"ohler, Agnes and Batt, Gregory and Hersen, Pascal , title =. Nature Communications , volume =

    work page

  21. [21]

    Physical Biology , volume =

    Jaruszewicz-Blo\'nska, Joanna and Lipniacki, Tomasz , title =. Physical Biology , volume =

    work page

  22. [22]

    Physical Review E , volume =

    Loinger, Adiel and Biham, Ofer , title =. Physical Review E , volume =

    work page

  23. [23]

    and Leibler, Stanislas , title =

    Elowitz, Michael B. and Leibler, Stanislas , title =. Nature , volume =

    work page

  24. [24]

    and Cantor, Charles R

    Gardner, Tim S. and Cantor, Charles R. and Collins, James J. , title =. Nature , volume =

    work page

  25. [25]

    and Dunlop, Mary J

    Lugagne, Jean-Baptiste and Blassick, Caroline M. and Dunlop, Mary J. , title =. Nature Communications , volume =

    work page

  26. [26]

    and Li, Xin and Duan, Cheng and Qian, Lin and Wu, Fan and Dougherty, Edward R

    Oduola, Wuraola O. and Li, Xin and Duan, Cheng and Qian, Lin and Wu, Fan and Dougherty, Edward R. , title =. Cancer Informatics , volume =

    work page

  27. [27]

    , title =

    Goodwin, Brian C. , title =

    work page

  28. [28]

    and Boada, Yadira and Pic\'o, J

    Pic\'o-Marco, E. and Boada, Yadira and Pic\'o, J. and Vignoni, A. , title =. Proceedings of the 6th IFAC Conference on Foundations of Systems Biology in Engineering , series =

    work page

  29. [29]

    Proceedings of the IEEE , volume =

    Khammash, Mustafa and Di Bernardo, Mario and Di Bernardo, Diego , title =. Proceedings of the IEEE , volume =

    work page

  30. [30]

    and P\'ajaro, Manuel and Szederk\'enyi, G\'abor and Otero-Muras, Irene , title =

    Fern\'andez, Christian and Faquir, Hamza and Vaghy, Mih\'aly A. and P\'ajaro, Manuel and Szederk\'enyi, G\'abor and Otero-Muras, Irene , title =. IFAC-PapersOnLine , volume =

    work page

  31. [31]

    IFAC-PapersOnLine , volume =

    Polcz, P\'eter and Szederk\'enyi, G\'abor , title =. IFAC-PapersOnLine , volume =

    work page

  32. [32]

    , title =

    Lohmiller, Winfried and Slotine, Jean-Jacques E. , title =. Automatisierungstechnik , volume =

    work page

  33. [33]

    Nature Biotechnology , volume =

    Milias-Argeitis, Andreas and Summers, Sean and Stewart-Ornstein, Jacob and Zuleta, Ignacio and Pincus, David and El-Samad, Hana and Khammash, Mustafa and Lygeros, John , title =. Nature Biotechnology , volume =

    work page

  34. [34]

    Nielsen, R. F. and Gernaey, K. V. and Mansouri, S. S. , title =. Computers & Chemical Engineering , volume =

    work page

  35. [35]

    and Zhang, Y

    Wang, L. and Zhang, Y. and Xu, M. and Liu, Q. and Wang, B. , title =. International Journal of Agricultural and Biological Engineering , volume =

    work page

  36. [36]

    and Huang, K

    Zhang, X. and Huang, K. W. and Vo, D.-N. and Han, M. and Decardi-Nelson, B. and Yin, X. , title =

    work page

  37. [37]

    Automatica , volume =

    Large Scale Model Predictive Control with Neural Networks and Primal Active Sets , author =. Automatica , volume =

    work page

  38. [38]

    Automatica , volume =

    Using stochastic programming to train neural network approximation of nonlinear Model Predictive Control laws , author =. Automatica , volume =

    work page

  39. [39]

    Automatica , volume =

    Menolascina, Filippo and Di Bernardo, Mario and Di Bernardo, Diego , title =. Automatica , volume =

    work page

  40. [40]

    An Optimal Switching Sequence Model Predictive Control Scheme for the

    Herrera, Felipe and Mora, Andrés and Cárdenas, Roberto and Díaz, Matías and Rodríguez, José and Rivera, Marco , journal =. An Optimal Switching Sequence Model Predictive Control Scheme for the

    work page

  41. [41]

    Automatica , volume =

    Finite control set model predictive control with limit cycle stability guarantees , author =. Automatica , volume =. 2025 , pages =

    work page 2025

  42. [42]

    Automatica , volume =

    Mosca, Edoardo , title =. Automatica , volume =

    work page

  43. [43]

    Automatica , volume =

    Wu, Tong and Zhang, Lixian and Lu, Shengao and Che, Weiming and Xu, Kaixin , title =. Automatica , volume =

    work page

  44. [44]

    Proceedings of the 49th IEEE Conference on Decision and Control (CDC) , year =

    Klavins, Eric , title =. Proceedings of the 49th IEEE Conference on Decision and Control (CDC) , year =

    work page

  45. [45]

    Proceedings of the 2023 European Control Conference (ECC) , year =

    Brancato, Sara Maria and De Lellis, Francesco and Salzano, Davide and Russo, Giovanni and di Bernardo, Mario , title =. Proceedings of the 2023 European Control Conference (ECC) , year =

    work page 2023

  46. [46]

    and Schaerli, Yolanda and Isalan, Mark and Briscoe, James and Page, Karen M

    Perez-Carrasco, Ruben and Barnes, Chris P. and Schaerli, Yolanda and Isalan, Mark and Briscoe, James and Page, Karen M. , title =. Cell Systems , volume =

    work page

  47. [47]

    ACS Synthetic Biology , volume =

    Guarino, Agostino and Fiore, Davide and Salzano, Davide and di Bernardo, Mario , title =. ACS Synthetic Biology , volume =

    work page

  48. [48]

    International Journal of Adaptive Control and Signal Processing , volume =

    Suzuki, Seiichi and Kawahara, Mutsuto , title =. International Journal of Adaptive Control and Signal Processing , volume =

    work page

  49. [49]

    and Otero-Muras, Irene , title =

    Sequeiros, Carlos and V\'azquez, Carlos and Banga, Julio R. and Otero-Muras, Irene , title =. Bioinformatics , volume =

    work page

  50. [50]

    Nature Communications , volume =

    Zhang, Fengyu and Sun, Yanhong and Zhang, Yihao and Shen, Wenting and Wang, Shujing and Ouyang, Qi and Luo, Chunxiong , title =. Nature Communications , volume =

    work page

  51. [51]

    Nature Communications , volume =

    Cannarsa, Maria Cristina and Liguori, Filippo and Pellicciotta, Nicola and Frangipane, Giacomo and Di Leonardo, Roberto , title =. Nature Communications , volume =

    work page

  52. [52]

    and Kutz, J

    Brunton, Steven L. and Kutz, J. Nathan , title =. 2019 , address =

    work page 2019

  53. [53]

    2016 , address =

    Goodfellow, Ian and Bengio, Yoshua and Courville, Aaron , title =. 2016 , address =

    work page 2016

  54. [54]

    IFAC-PapersOnLine , volume =

    Fern\'andez, Christian and Faquir, Hamza and P\'ajaro, Manuel and Otero-Muras, Irene , title =. IFAC-PapersOnLine , volume =

    work page

  55. [55]

    , title =

    Del Vecchio, Domitilla and Murray, Richard M. , title =. 2014 , address =

    work page 2014

  56. [56]

    Ale, Aleksandar and Kirk, Paul and Stumpf, Michael P. H. , title =. The Journal of Chemical Physics , volume =

    work page

  57. [57]

    and Hellander, Andreas and Petzold, Linda R

    Gillespie, Daniel T. and Hellander, Andreas and Petzold, Linda R. , title =. The Journal of Chemical Physics , volume =

    work page

  58. [58]

    and Kurtz, Thomas G

    Anderson, David F. and Kurtz, Thomas G. , title =. Springer , year =

    work page

  59. [59]

    , title =

    Pavliotis, Grigorios A. , title =

    work page

  60. [60]

    , title =

    Gardiner, Crispin W. , title =

    work page

  61. [61]

    2022 , author =

    Interval analysis of worst-case stationary moments for stochastic chemical reactions with uncertain parameters , journal =. 2022 , author =

    work page 2022

  62. [62]

    2016 , author =

    Contraction analysis of switched systems via regularization , journal =. 2016 , author =

    work page 2016

  63. [63]

    Event-triggered PDF shape control of non-Gaussian stochastic system , journal =

    Xiaoyang Sun and Ping Zhou , volume =. Event-triggered PDF shape control of non-Gaussian stochastic system , journal =

    work page

  64. [64]

    and Gupta, Ankit and Khammash, Mustafa , booktitle=

    Zand, Armin M. and Gupta, Ankit and Khammash, Mustafa , booktitle=. Control with Practical Guarantees of Stationary Variance in Stochastic Chemical Reaction Networks , year=

    work page