Dynamical Systems · arXiv:2611.17232 v1 · submitted 12 Nov 2026 · 22 references
Luenberger observers for two-gene regulatory networks with sigmoidal regulation: explicit observability conditioning, polytopic convergence certificates, and a logistic–Hill comparison
The paper says
“Only a few species of a gene regulatory network can be measured, so the others must be reconstructed by a state observer, and how well this works depends on the function chosen to describe regulation.”
more of the abstract
We examine this for the two-gene negative feedback loop with one measured protein, for monotone regulation functions including the logistic function and the Hill function with its leaky and shifted variants. Using standard tools, we show that the answer is set by the smallest and the largest slope of the regulation function on the concentrations visited. For every function considered, the unmeasured concentration can be recovered from the measured signal and its rate of change, with an error sensitivity set by the smallest slope. A Luenberger observer with explicit gains converges exponentially from any initial guess and tolerates bounded measurement noise whenever a checkable matrix condition involving only the largest slope holds, as it does for every model examined with bounded slopes. The guaranteed rate cannot exceed the degradation rate of the unmeasured protein unless the smallest slope is positive, and faster rates then hold only for small initial errors. All reported rates are certified from above and below. At matched threshold and steepness, no family is uniformly better: the logistic function is better conditioned when the threshold lies mid-range, leaky or shifted Hill functions when it lies low. Hill functions with exponent below one have an unbounded slope at zero concentration and lose the global guarantee, which logistic and shifted Hill functions keep.
from the abstract
Pith says
Machine review · deepseek-v4-flash · 13 Nov 2026 · what was read was not recorded · not checked by a person
It can be wrong. Challenge this review or reply to it.
Holds up. A careful, honest math paper that backs its central slope-based claim with proofs and reproducible certificates; send it to review. This line is wrong
4 minor
How often objections from reviews like this one are right: not yet measured
Not checked: figures, code, what the 22 references say.
4 minor objections
The paper says, at Section 4.5 and Appendix B (Table B.1)
Pith says: minor objection 1 of 4
The upper-side certificates are not fully self-contained: for the rows marked "dual" only lambda_min(G) is printed, while the factors Lv of the certificate matrices are said to be in certificates_data.txt, available from the author upon request. Because the paper advertises that all reported rates are certified from above and below, please include the certificate data file as supplementary material or in an archived repository so that the dual certificates can be recomputed from the printed factors.
The paper says, at Section 2.3 (proof of Lemma 1)
Pith says: minor objection 2 of 4
The symbols used for the liminf and the limsup of xi(t) are typographically almost identical, which makes the two displayed inequalities in the proof hard to distinguish; please use visually distinct notation (for example x with a lower bar and x with an upper bar) and state explicitly which inequality bounds the limsup and which bounds the liminf.
show 2 more
The paper says, at Section 4.4 (Eq. (15))
Pith says: minor objection 3 of 4
The closed form eta(c1,c2)=T/2 - sqrt(max{T^2/4 - D(c1,c2), 0}) silently covers the complex-root case T^2 < 4D, in which eta = T/2; a one-sentence explanation of the two cases would prevent a misreading of the ceiling and of the equality statement for l2 <= kappa2*m2.
The paper says, at Abstract and Section 6.3
Pith says: minor objection 4 of 4
The abstract's dichotomy ("the logistic function is better conditioned when the threshold lies mid-range, leaky or shifted Hill functions when it lies low") is drawn from the two scenarios A and B of Section 6.1 together with the tail-decay discussion in Section 6.3(3), where the extrapolation is already hedged with "roughly"; consider carrying the same hedge into the abstract or stating explicitly that the dichotomy is scenario-based.
What would settle it
Verify that the four-vertex matrix inequality (11) is feasible for the logistic model with the baseline parameters and the gains given in Section 4.1, using the matrix P printed in (17). This is already done in the paper (residual -0.019). A reader could independently recompute this check with any LMI solver to confirm the certificate.
from the skeptic's reading
How the review went
4 readings by deepseek-v4-flash · 13 Nov 2026
- Reader holds up
- Skeptic agrees
- Editor wrote the verdict above
- Referee minor revision: 4 minor
Not run on this paper: the simulated authors' defence.
Pith did not check this review's objections against the paper.
The review in full
The review in brief, machine-written
Quick take: this is a careful, honest mathematics paper on observer design for a two-gene negative feedback loop, comparing logistic and Hill-type regulation functions. The central claim holds: observability conditioning and guaranteed observer convergence are set by the smallest and largest slopes of the regulation functions on the relevant set. I would send it to peer review.
What is actually new: not the tools — Luenberger observers, the differential mean-value error representation, and four-vertex LMIs are standard — but the systematic attribution of each conclusion to a specific slope property, and an explicit logistic-versus-Hill comparison for observer design. The leaky and shifted Hill variants are handled carefully, and the separate treatment of Hill exponents n<1, where the upper slope bound is lost at zero, is honest and technically sound. The paper does not oversell: it says clearly that it proposes no new observer methodology and that the matching of families at one point is a convention, not a fit.
The certification practice is a real strength. Appendix B prints the Lyapunov matrices and residuals for every reported rate, brackets the optimal rate within 10−4, and supplies dual certificates for the upper bounds. The numerical claims are reproducible in a way that is uncommon in this literature.
Soft spots, in proportion. First, the quantitative comparison evaluates slopes on the invariant box B∞, which is an outer enclosure rather than the set of concentrations actually visited. For the baseline parameters the trajectories converge to a stable equilibrium, so the minimal slopes at the ends of B∞ are worst-case, not typical. The paper states that B∞ is an enclosure, so this is not a hidden flaw, but the numerical margins (for example Hill m1 around 10−22) should not be read as representative of operating behaviour. Second, the comparison rests on two parameter scenarios, mid-range and low threshold; the qualitative tail-decay argument is sound, but the ranking is not proven to hold beyond those examples. Third, the global certificates are conservative (noise gain 53.4W versus the local bound 2.7W) and the fast set-based rates come with very small admissible initial errors; both are acknowledged in the text. Fourth, the code and certificate data are available 'upon request' rather than in a stable repository, which is a minor reproducibility blemish.
I found no load-bearing flaw. The proofs check, the assumptions are stated, and the limitations are laid out by the author. This is a solid contribution for people choosing between Hill and logistic regulation models in observer-based monitoring of gene circuits.
The strongest objection
The paper is a rigorous, self-contained mathematical analysis of observability and Luenberger observer design for a specific two-gene negative feedback loop. The central claim—that the smallest and largest slopes of the regulation function determine observability and observer convergence—is supported by explicit theorems (Theorem 1, Theorem 2, Proposition 3) that are proved from stated assumptions (Assumption 1). The paper carefully delineates its scope (Section 2.2, Section 7), acknowledges limitations (Section 8), and provides numerical certificates for all reported rates (Appendix B). The weakest assumption identified by the reader—that regulation functions are C1 with bounded slopes—is indeed standard for the tools used and is explicitly discussed; the paper treats the case n<1 separately where this fails. I find no internal inconsistency, hidden assumption, or unsupported leap in the central argument. The comparison between families is based on computed quantities and is clearly qualified. The paper does what it sets out to do, and the central claim holds under the stated conditions.
The reader's verdict is ACCEPT with high confidence, and the weakest assumption is that the regulation functions are C1 with bounded slopes. I agree that this is the weakest link in the sense that it is the most restrictive condition, but it is a standard assumption for the tools used (Luenberger observer, LMI certificates) and the paper explicitly addresses its limitations: Section 7 treats Hill exponents n<1 where the assumption fails, and the paper notes that for n<1 only set-based certificates exist. The paper also acknowledges that the observer assumes known parameters and that robustness is not analysed. The central claim is well-supported by the theorems and numerical examples. I see no load-bearing concern that would warrant changing the verdict. The paper is a solid contribution within its stated scope.
The referee's report
The paper is rigorous, self-contained, and carefully scoped. Each central conclusion is proved from the stated assumptions and explicitly attributed to the slope property it uses: injectivity of r1 for global observability, the smallest slope m1 for conditioning, the largest slope for the LMI certificate, and the sign of m1 for the rate ceiling. Strengths worth emphasising: proofs appear in the text for all central statements (Theorems 1-3, Propositions 2-3, Lemmas 1-3); the family comparison is computed from explicit formulas under a stated matching convention rather than fitted; the certification methodology of Section 4.5 and Appendix B brackets every reported rate from below with a printed matrix and its residual and from above either by the closed-form ceiling or by an explicit dual certificate; and limitations (known parameters, no delays, the n<=1 irregularity, case-by-case verification of LMI feasibility) are acknowledged in Sections 2.2, 7 and 8.
Pith's reading of the paper (a machine's summary, not the authors')
This paper examines how the choice of sigmoidal regulation function—logistic, Hill, leaky Hill, or shifted Hill—affects the ability to reconstruct unmeasured protein concentrations from a single measured output in a two-gene negative feedback loop. The central claim is that the smallest and largest slopes of the regulation function on the operating set determine observability conditioning and the convergence guarantees of a Luenberger observer. The paper provides explicit formulas and certificates for all families, showing that the logistic function is better conditioned when the threshold lies mid-range, while leaky or shifted Hill functions are better when it lies low, but that certified convergence rates differ little between families.
What carries the argument. The key machinery is the mean-value (polytopic) representation of the observer error dynamics, which expresses the error system as a linear parameter-varying system with coefficients confined to a rectangle of slopes determined by the minimum and maximum slopes of the regulation functions on the operating set. This reduces the convergence problem to a four-vertex linear matrix inequality (LMI) feasibility problem. The observability analysis uses the Lie-derivative map and its singular values, with the smallest singular value determined by the smallest slope.
Core claim. The paper establishes that for the two-gene negative feedback loop with one measured protein, the observability map's conditioning and the exponential convergence rate of a Luenberger observer are set by the smallest and largest slope of the regulation function on the concentrations visited. The unmeasured concentration can be recovered from the measured signal and its rate of change, with error sensitivity inversely proportional to the smallest slope. A Luenberger observer with explicit gains converges exponentially from any initial guess and tolerates bounded measurement noise whenever a checkable matrix condition involving only the largest slope holds. The guaranteed rate cannot exceed the degradation rate of the unmeasured protein unless the smallest slope is positive. All reported rates are certified from above and below using polytopic matrix inequalities.
The premise it rests on. The regulation functions r1 and r2 are C1 and have bounded slopes on R (Assumption 1), which excludes discontinuities, non-smooth regulation, and Hill functions with exponent below one (these are treated separately but lose the global guarantee).
If the paper is right (Pith's inferences, not the authors')
The explicit conditioning formulas allow practitioners to compare regulation families before fitting data, by computing the smallest slope on the expected operating range.
The four-vertex LMI certificate provides a checkable condition for global exponential convergence of a Luenberger observer for any sigmoidal regulation function with bounded slopes.
The ceiling on the certified rate (Proposition 3) shows that without a positive lower slope bound, the guaranteed convergence rate cannot exceed the degradation rate of the unmeasured protein, which is a fundamental limitation.
The comparison shows that the logistic function is not uniformly better than Hill-type functions; the ranking depends on where the threshold lies relative to the expression range.
For Hill exponents below one, the unbounded slope at zero concentration prevents a global certificate, but set-based certificates on intervals bounded away from zero are still feasible.
The paper's framework could be extended to longer regulatory cascades, where the product of slopes along the chain would compound the conditioning (as shown in Appendix A).
The dependence on the smallest slope suggests that regulation functions with a strictly positive basal slope (like logistic or shifted Hill) are inherently more robust for observer design near the basal expression level.
The explicit comparison between families provides a rational basis for choosing a regulation function in observer-based control or monitoring of synthetic gene circuits, beyond the convention of using Hill functions.
The polytopic certificate method could be applied to other nonlinear systems where the nonlinearity is monotone and its slope range is known, not just gene regulation.
Check this review
Pith stored no record of this review's passes and took no fingerprint of it. Reviews made by code from 13 Nov 2026 on carry both.
How often reviews like this one are right
Reviews like this one: written by deepseek-v4-flash, where what was read was not recorded.
How often objections from reviews like this one are right: not yet measured
Reviews by people
This review was written by a machine. No person has reviewed this paper on Pith yet. A scientist listed on Pith can review it for a fee, and their review is published here whatever it concludes.
The paper
Abstract
Only a few species of a gene regulatory network can be measured, so the others must be reconstructed by a state observer, and how well this works depends on the function chosen to describe regulation. We examine this for the two-gene negative feedback loop with one measured protein, for monotone regulation functions including the logistic function and the Hill function with its leaky and shifted variants. Using standard tools, we show that the answer is set by the smallest and the largest slope of the regulation function on the concentrations visited. For every function considered, the unmeasured concentration can be recovered from the measured signal and its rate of change, with an error sensitivity set by the smallest slope. A Luenberger observer with explicit gains converges exponentially from any initial guess and tolerates bounded measurement noise whenever a checkable matrix condition involving only the largest slope holds, as it does for every model examined with bounded slopes. The guaranteed rate cannot exceed the degradation rate of the unmeasured protein unless the smallest slope is positive, and faster rates then hold only for small initial errors. All reported rates are certified from above and below. At matched threshold and steepness, no family is uniformly better: the logistic function is better conditioned when the threshold lies mid-range, leaky or shifted Hill functions when it lies low. Hill functions with exponent below one have an unbounded slope at zero concentration and lose the global guarantee, which logistic and shifted Hill functions keep.
The paper's own sentences, arranged, kept beside the review and separate from it. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. No sentence in this record was written by a model.
abstract compiled · 2026-11-14
The authors' main claim, as the classifier picked it out 0.53
“Using standard tools, we show that the answer is set by the smallest and the largest slope of the regulation function on the concentrations visited.”
How the main claim is stated: asserted 1.00
What the abstract itself states, as probabilities: claims first 0.03 · states limitation 0.92 · human participants 0.02 · controlled comparison 0.12 · randomized assignment 0.01 · states quantitative result 0.78 · compares with prior results 0.59 · mentions code or data availability 0.02
The abstract, sentence by sentence
This table labels the abstract only (arXiv listing abstract v1). The number is the probability the classifier gave the label.
| Sentence | Role | The authors' own work | States the main conclusion | Strength |
|---|---|---|---|---|
| Only a few species of a gene regulatory network can be measured, so the others must be reconstructed by a state observer, and how well this works depends on the function chosen to describe regulation. | background 1.00 | 0.05 | 0.05 | asserted 1.00 |
| We examine this for the two-gene negative feedback loop with one measured protein, for monotone regulation functions including the logistic function and the Hill function with its leaky and shifted variants. | contribution 0.64 | 0.72 | 0.04 | asserted 0.85 |
| Using standard tools, we show that the answer is set by the smallest and the largest slope of the regulation function on the concentrations visited. | result 0.96 | 0.90 | 0.45 | asserted 1.00 |
| For every function considered, the unmeasured concentration can be recovered from the measured signal and its rate of change, with an error sensitivity set by the smallest slope. | result 1.00 | 0.87 | 0.16 | asserted 1.00 |
| A Luenberger observer with explicit gains converges exponentially from any initial guess and tolerates bounded measurement noise whenever a checkable matrix condition involving only the largest slope holds, as it does for every model examined with bounded slopes. | result 0.86 | 0.91 | 0.44 | asserted 1.00 |
| The guaranteed rate cannot exceed the degradation rate of the unmeasured protein unless the smallest slope is positive, and faster rates then hold only for small initial errors. | result 0.81 | 0.88 | 0.11 | asserted 0.98 |
| All reported rates are certified from above and below. | result 0.54 | 0.89 | 0.05 | asserted 1.00 |
| At matched threshold and steepness, no family is uniformly better: the logistic function is better conditioned when the threshold lies mid-range, leaky or shifted Hill functions when it lies low. | result 0.98 | 0.88 | 0.20 | asserted 1.00 |
| Hill functions with exponent below one have an unbounded slope at zero concentration and lose the global guarantee, which logistic and shifted Hill functions keep. | result 0.97 | 0.87 | 0.09 | asserted 1.00 |
What kind of paper the classifier reports this is: theoretical result 1.00 · evidence from proof or derivation 1.00
What was examined, and what was not
Sources read
- arXiv listing abstract v1
Not examined
- full text
- figures and tables
- references and citing sentences
- relations to other papers
Levels run
- P abstract
Not run
- S sections
- T body sentences
- C citations
- R cross-paper relations
How the labels were made
Compiled by classifier-1.13.0 via TypeSafe API, question bank abstract-p1-20260923, 2026-11-14: 1 request, 7,009 input tokens.
Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. This record as JSON
Figures (2)
Figures are the authors', reproduced from the paper.
References (22)
- [1] U. Alon, An Introduction to Systems Biology: De- sign Principles of Biological Circuits, Chapman and Hall/CRC, Boca Raton, FL, 2007 on Pith
- [2] H. de Jong, Modeling and simulation of genetic regu- latory systems: a literature review, Journal of Compu- tational Biology 9 (1) (2002) 67–103.doi:10.1089/ 10665270252833208 on Pith
- [3] A. V . Hill, The possible effects of the aggregation of the molecules of haemoglobin on its dissociation curves, Journal of Physiology 40 (1910) iv–vii, supplement (Pro- ceedings of the Physiological Society) on Pith
- [4] I. Belgacem, Logistic gene regulatory networks: A model- ing framework beyond Hill functions, Mathematical Bio- sciences 401 (2026) 109803.doi:10.1016/j.mbs. 2026.109803 on Pithdoi:10.1016/j.mbs
- [5] R. Hermann, A. J. Krener, Nonlinear controllability and observability, IEEE Transactions on Automatic Con- trol 22 (5) (1977) 728–740.doi:10.1109/TAC.1977. 1101601 on Pithdoi:10.1109/tac.1977
- [6] G. Besançon (Ed.), Nonlinear Observers and Applica- tions, V ol. 363 of Lecture Notes in Control and Infor- mation Sciences, Springer, Berlin, 2007.doi:10.1007/ 978-3-540-73503-8 on Pith
- [7] D. G. Luenberger, An introduction to observers, IEEE Transactions on Automatic Control 16 (6) (1971) 596– 602.doi:10.1109/TAC.1971.1099826 on Pithdoi:10.1109/tac.1971.1099826
- [8] R. Rajamani, Observers for Lipschitz nonlinear systems, IEEE Transactions on Automatic Control 43 (3) (1998) 397–401.doi:10.1109/9.661604 on Pithdoi:10.1109/9.661604
- [9] A. Zemouche, M. Boutayeb, On LMI conditions to design observers for Lipschitz nonlinear systems, Automatica 49 (2) (2013) 585–591.doi:10.1016/j.automatica. 2012.11.029 on Pithdoi:10.1016/j.automatica
- [10] M. Arcak, P. Kokotovi´c, Nonlinear observers: a circle cri- terion design and robustness analysis, Automatica 37 (12) (2001) 1923–1930.doi:10.1016/S0005-1098(01) 00160-1 on Pithdoi:10.1016/s0005-1098(01
- [12] J.-P. Gauthier, H. Hammouri, S. Othman, A simple ob- server for nonlinear systems applications to bioreactors, IEEE Transactions on Automatic Control 37 (6) (1992) 875–880.doi:10.1109/9.256352 on Pithdoi:10.1109/9.256352
- [13] E. D. Sontag, Molecular systems biology and control, European Journal of Control 11 (4–5) (2005) 396–435. doi:10.3166/ejc.11.396-435 on Pithdoi:10.3166/ejc.11.396-435
- [14] D. Bouhadjra, A. Alessandri, P. Bagnerini, A. Zemouche, High-gain estimation of mRNA and protein concentra- tions of a genetic regulatory network, in: European Con- trol Conference (ECC), IEEE, London, United Kingdom, 2022, pp. 578–583 on Pith
- [15] A. V . Proskurnikov, A. Medvedev, A simple positive state observer for multidimensional Goodwin’s oscillator, in: 18th European Control Conference (ECC), IEEE, Naples, Italy, 2019, pp. 1671–1676.doi:10.23919/ECC.2019. 8795718 on Pithdoi:10.23919/ecc.2019
- [16] M. Reza-Rahmati, G. Flores, A nonlinear observer for bi- linear systems in block form, European Journal of Con- trol 70 (2023) 100780.doi:10.1016/j.ejcon.2023. 100780 on Pithdoi:10.1016/j.ejcon.2023
- [17] L. Chambon, I. Belgacem, J.-L. Gouzé, Qualitative con- trol of undesired oscillations in a genetic negative feed- back loop with uncertain measurements, Automatica 112 (2020) 108642.doi:10.1016/j.automatica.2019. 108642 on Pithdoi:10.1016/j.automatica.2019
- [18] I. Belgacem, J.-L. Gouzé, R. Edwards, Control of nega- tive feedback loops in genetic networks, in: 59th IEEE Conference on Decision and Control (CDC), IEEE, Jeju Island, Republic of Korea, 2020, pp. 5098–5105.doi: 10.1109/CDC42340.2020.9304088 on Pithdoi:10.1109/cdc42340.2020.9304088
- [19] B. C. Goodwin, Oscillatory behavior in enzymatic control processes, Advances in Enzyme Regulation 3 (1965) 425– 437.doi:10.1016/0065-2571(65)90067-1 on Pithdoi:10.1016/0065-2571(65)90067-1
- [20] M. B. Elowitz, S. Leibler, A synthetic oscillatory network of transcriptional regulators, Nature 403 (6767) (2000) 335–338.doi:10.1038/35002125 on Pithdoi:10.1038/35002125
- [21] T. S. Gardner, C. R. Cantor, J. J. Collins, Construction of a genetic toggle switch in Escherichia coli, Nature 403 (6767) (2000) 339–342.doi:10.1038/35002131. 12 Table B.1: Certificates for the reported rates (min −1). Labels A, B, C refer to the scenarios of Sections 6 and 7... on Pithdoi:10.1038/35002131
show 2 more
- [22] H. K. Khalil, High-Gain Observers in Nonlinear Feedback Control, Advances in Design and Control, SIAM, Philadelphia, PA, 2017.doi:10.1137/1. 9781611974867 on Pithdoi:10.1137/1
- [23] H. K. Khalil, Nonlinear Systems, 3rd Edition, Prentice Hall, Upper Saddle River, NJ, 2002. 13 on Pith
Challenges and discussion
Disagree with the review, or know something it missed? Leave a comment, or challenge this review formally: a challenge becomes a permanent entry in this paper's record, and its decision is published here.
Come back by email
A short list of the week's newly reviewed papers. Nothing is sent until you confirm from your inbox, and every email has a link that stops it. No account.
No comments yet.
Sign in to comment