Robust oscillations in multi-cyclic models of biochemical clocks

Clara del Junco; Suriyanarayanan Vaikuntanathan

REVIEW 3 major objections 4 minor 22 references

Robust oscillations in multi-cyclic models of biochemical clocks

T0 review · 3 major / 4 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The paper argues that high chemical affinity makes the period and coherence of multi-cycle Markov oscillators insensitive to the arrangement of cycles and rates, so a coarse-grained single-cycle theory predicts them accurately.

desk verdict A solid, honest extension of the authors' single-cycle theory, but the central robustness claim rests on an unproven two-moment truncation that is only tested in a narrow regime. read the letter →

arxiv 1909.02534 v2 pith:EXAYGK4Y submitted 2019-09-05 cond-mat.stat-mech cond-mat.softphysics.bio-ph

Clara del Junco , Suriyanarayanan Vaikuntanathan This is my paper · ORCID

classification cond-mat.stat-mechcond-mat.softphysics.bio-ph

keywords biochemicalclocksMarkovstatemodelsmulti-cyclenetworksfirst-passage-timecoarse-grainingoscillationperiodcoherenceenergydissipationinputcompensation

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

Biochemical clocks can be modelled as a large cycle of states with small side loops attached, representing the different paths a molecule may sample while completing one tick. This paper argues that when such a clock is driven far from equilibrium, its period and its phase coherence are determined by a much simpler object: a single cycle in which each side loop is replaced by one effective link. The replacement is made by matching the mean and variance of the first-passage-time distribution across each loop, and the resulting single-cycle network is then handled by an earlier analytical theory whose predictions depend on rate values but not on their positions. The authors verify numerically, including in networks with random loop shapes, positions, and rates, that this coarse-grained theory matches exact diagonalization at high chemical affinity. The consequence, if the claim holds, is that a large energy budget makes biochemical clocks robust to rewiring or rearrangements of their reaction network, because many different topologies become degenerate in the observables that matter for timekeeping.

What carries the argument

The argument rests on two devices. The first is first-passage-time coarse-graining: for each small side cycle, called a decoration, the Laplace-transformed first-passage-time distribution from a state upstream to a state downstream of the decoration is compared with the corresponding distribution for a line segment carrying two unknown rates, and the effective rates $\eta^+$ and $\eta^-$ are chosen to match the mean and variance. The second is the transfer-matrix theory for single-cycle rings, in which the relevant eigenvalue is written $\varphi = \varphi^{(0)} + C\gamma$ and $\gamma$ solves a self-consistent equation whose terms add separately for each nonuniform rate. Neither device needs to know the positions of the decorations or rates on the ring, which is precisely why the resulting predictions imply positional and topological robustness. The validity of the coarse-graining is deliberately restricted to small decorations, up to six sides, because beyond that the effective rates diverge or can become negative and a decoration can sustain coherent oscillations of its own.

What would settle it

Compute, by exact diagonalization, the period and coherence of a large decorated network whose decoration is a cis hexagon or larger, set the entry probability near the value at which the effective rates in Table I diverge, and drive at high affinity; if the exact T and R separate measurably from the coarse-grained single-cycle prediction, the two-moment mapping is not sufficient. The paper's own restriction to decorations of six sides or fewer marks where this test should start.

Watch

Extended reading notes

Core claim

The paper's central claim is that at high affinity $A$ the period $T$ and coherence $R$ of a multi-cycle Markov oscillator are set, to good accuracy, by a coarse-grained single-cycle network in which each decoration contributes two effective hopping rates obtained from the first-passage-time distribution across it. Since the analytical expression for the slowest eigenvalue $\varphi$ used to define $T$ and $R$ contains the nonuniform rates only through additive terms and carries no information about their relative positions, a successful prediction implies that $T$ and $R$ are insensitive to the arrangement of the cycles and the rates. The paper verifies the claim by exact numerical diagonalization of decorated networks, by diagonalization of the coarse-grained network, and by the analytical theory, for networks with symmetric placements, random decoration shapes and locations, and combined random rates, with agreement improving as the affinity increases. It also derives a linear compensation mechanism: when the affinity changes, tuning the probability of entering decorations in proportion to the change keeps the period nearly constant, compensating period shifts of about 50 percent down to deviations below 5 percent.

Load-bearing premise

The whole argument rests on the assumption that each small side loop's effect on the global clock is fully captured by the mean and variance of the time it takes to cross that loop, so all details of what happens inside the loop can be discarded.

Editorial extensions

If this is right

At high affinity, the period and coherence of a decorated oscillator can be computed from the distributions of rates and decorations alone, without knowing which edge a rate sits on or where a loop attaches.
Randomizing the shapes and positions of side loops, or adding quenched disorder to the rates, changes $T$ and $R$ only weakly when the chemical affinity is large; the scatter of exact, coarse-grained, and theoretical results collapses as $A$ increases.
The same coarse-graining works as a practical reduction: a multi-cycle network with many decorations can be replaced by a single-cycle network with effective rates, dramatically reducing the dimensionality of the calculation.
A one-parameter linear feedback, changing the probability of entering a decoration in proportion to a change in affinity, can hold the period fixed against changes in driving, with compensation becoming increasingly effective at high affinity and large main-cycle size.
The restriction to small decorations is a real limit: for loops with more than about six vertices, the effective rates diverge at small entry probabilities, so the theory does not claim to describe competing oscillatory modes.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

Editorial inference: the first-passage-time moment-matching coarse-graining should apply beyond circadian-style clocks to any Markov process consisting of a dominant cycle with attached metastable branches, such as molecular motors or stochastic enzyme cycles; a test would be to compare exact and coarse-grained period and coherence in such models.
Editorial inference: the linear compensation rule suggests a plausible evolutionary route to input compensation, because selection would need to tune only one parameter, the dwell probability in side reactions, rather than many rates, so robust period control could arise from a single mutation; the paper does not claim this.
Editorial inference: if the theory is right, a reconstituted in vitro oscillator experiment should show that increasing ATP concentration makes the period insensitive to mutations or perturbations that relabel reaction pathways while preserving average rates; the paper does not report such an experiment.
Editorial inference: the divergence of effective rates for larger decorations hints at a possible transition in oscillator behavior, where a decorated loop becomes its own clock and yields a multi-period regime; the paper flags this only as a boundary of its method.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The headline is simple: this is a reasonable extension of the authors' own single-cycle theory to decorated multi-cycle networks, and the numerical agreement with exact diagonalization is good in the regime they actually test. What is new is the first-passage-time moment-matching coarse-graining that maps each small cycle to an effective link, plus the demonstration that period and coherence become insensitive to the arrangement of these decorations as affinity increases. The coarse-graining is parameter-free, which is a real virtue, and the authors are unusually candid about where it fails: effective rates diverge or go negative for larger decorations, and they restrict themselves to small cycles with cis-configured rates.

The soft spot is exactly the one the stress-test note flags. Matching the mean and variance of the local first-passage-time distribution does not, by itself, guarantee that the global dominant eigenvalue—which controls T and R—is preserved. Higher FPT cumulants, or internal dynamics of the decoration that oscillate on their own, could in principle matter. The paper offers no argument, only the empirical evidence from the tested cases. Because the scatter plots are normalized per panel and no quantitative error metric is given, it is hard to tell when the approximation degrades.

I think the central claim—that high energy dissipation makes these observables insensitive to network arrangement—is plausible but not proven in the generality the conclusion implies. What is proven is narrower: for small decorations (3-6 sides) with cis rates, modest µ, and high affinity, the two-moment coarse-graining works. That is a useful, honest result.

The paper deserves serious peer review. I would send it, but I'd ask the authors to either strengthen the justification of the moment truncation (e.g., perturbation analysis showing higher cumulants decouple at high affinity) or explicitly reframe the conclusion to the tested class. The input compensation example is a nice bonus but not central.

Who this is for: people working on stochastic thermodynamics of clocks and possibly synthetic oscillator design. It's not a landmark, but it's a solid incremental contribution.

Referee Report

3 major / 4 minor

Summary. The paper extends a previously developed single-cycle theory of Markov-state biochemical oscillators to networks consisting of a large cycle decorated with small secondary cycles. The decoration is coarse-grained into two effective transition rates matched to the mean and variance of the first-passage-time (FPT) distribution across the decoration, reducing the multi-cycle network to a single-cycle network. The authors then apply their earlier high-affinity analytical theory, which depends on the rates but not on their spatial arrangement, to predict the oscillation period T and coherence R. They compare these predictions with exact numerical diagonalization of the full and coarse-grained networks, including networks with random topology and random rates, and report agreement at high chemical affinity. The central conclusion is that high energy dissipation makes T and R insensitive to the arrangement of cycles and rates; the paper also demonstrates a linear-compensation mechanism in which tuning the decoration entry probability maintains a constant period when the affinity changes.

Significance. If the main claim holds, the paper offers a parameter-free coarse-graining that maps a class of multi-cyclic Markov models onto single-cycle theory, with the practical implication that biochemical clocks can become robust to topological and rate fluctuations when sufficiently driven. The study is carefully validated against exact numerical diagonalization for the tested class, and the coarse-graining introduces no fitted parameters. The explicit analytical formulas for effective rates (Table I and Appendix C) and the modification of the earlier perturbation replacement (Appendix A) are useful contributions. The demonstration of input compensation via the decoration parameter is a nice additional result. However, the central theoretical justification for matching only the first two FPT moments is not established, which limits the strength of the general robustness claim.

major comments (3)

[Section IV, Eq. (8)] The mapping from a decoration to a single link matches only the mean and variance of the first-passage-time (FPT) distribution; no argument is provided that the dominant eigenvalue φ, and hence T and R, is a function solely of these two moments. Because the central claim of insensitivity to network arrangement rests on the predictive accuracy of this two-moment truncation, the manuscript needs either a derivation in the high-affinity limit showing that higher FPT cumulants enter only at subleading order, or a quantitative test comparing networks with identical FPT mean and variance but different higher moments. The paper's own caveat (Section IV, paragraph after Table I) that effective rates diverge and that larger decorations can support their own oscillations underscores that the truncation is uncontrolled in general.
[Appendix A, Eqs. (A5)-(A6)] The replacement X(0)_1 -> X(1)_1 is introduced ad hoc, with the statement that the previous cancellation of terms no longer holds when h-/h+ > 1. The validity of this replacement is not derived; it is only validated by the numerical comparisons. Since all theory curves in Figs. 3-5 rely on this modified perturbation theory, the manuscript should provide a derivation or an error estimate showing that the replacement preserves the eigenvalue to the same order in k-/k+ and in the coarse-graining parameters.
[Section V, Figs. 4 and 5] The numerical evidence supporting the main claim is limited to small decorations (3-6 sides), the cis rate configuration (with trans discussed only briefly), and non-overlapping decorations separated by at least one edge; moreover, the scatter plots are normalized per panel, which can mask systematic deviations, and no quantitative error metric is reported. A stronger test of the robustness claim would include larger decorations (where effective rates remain positive), overlapping decorations, and plots of absolute or relative errors as a function of affinity.

minor comments (4)

[Section II, Eq. (1)] The affinity is defined as A = ∑_cycle k+_i/k-_i, which is dimensionally inconsistent and should be A = ∑_cycle ln(k+_i/k-_i); as printed, the subsequent statements involving exp(A/N) do not follow.
[Section VI and Fig. 6 caption] The text states that the period with fixed μ is shown as solid lines in Fig. 6, while the Fig. 6 caption says solid lines are the compensated period with μ varied; one of the two is incorrect.
[Appendix B, Eq. (B1)] The symbol μ is used both for the decoration entry probability (Fig. 1c and throughout) and for a time increment (Eq. B1 writes t1 = t0 + μt); this notational collision should be removed by using Δt for the time increment.
[Figure 4 and Figure 5 captions] Stating that all values are normalized by the largest value in each scatter plot obscures the absolute magnitude of deviations; displaying unnormalized values or difference plots would make the convergence quantitative.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the coarse-graining matches local first-passage-time moments, the prior analytic theory is independent and parameter-free, and all predictions are checked against exact full-network numerics.

full rationale

The paper's derivation chain is self-contained rather than circular. The analytic period/coherence expressions (Eqs. 4-6) are taken from the authors' prior Ref. 12, but that theory is a parameter-free, independently published eigenvalue calculation for single-cycle networks and does not contain the multi-cycle robustness result as an input. The coarse-graining step in Section IV determines the effective rates η± by equating the mean and variance of the first-passage-time distribution across each decoration with those of a two-rate link (Eq. 8 and Appendix B); this is a local, analytic matching condition that involves no fitted parameters and never uses the target observables T or R as inputs. The subsequent prediction of T and R from the coarse-grained rates is a nontrivial application of the single-cycle eigenvalue theory, and the paper validates it against exact numerical diagonalization of the full multi-cycle transition matrix (Figs. 3-5). The acknowledged limitation that higher FPT moments or internal decoration dynamics could invalidate the two-moment truncation is an approximation risk, not a circularity: the reduction is approximate and explicitly tested, not assumed by definition. The self-citation to Ref. 12 is load-bearing but independent under the stated criteria, so it does not raise the circularity score.

Assumptions & free parameters 0 free parameters · 6 assumptions · 0 invented entities

The paper fits no parameters to data: all rates are model inputs, and the effective rates eta+ and eta- are determined by matching two moments of the first-passage time distribution. No new physical entities are postulated. The load-bearing content is carried by the assumptions listed above; the most fragile are the first-passage-time moment truncation and the X(1) perturbation replacement.

assumptions (6)

domain assumption Oscillator dynamics can be represented by a Markov state model on a graph; rates along edges encode the underlying chemical feedback and mass action kinetics.
Section II explicitly calls the model an 'emergent picture' rather than the chemical reaction network, so the mapping from chemistry to rates is assumed.
domain assumption High-affinity limit: terms of order k-/k+ = exp(-A/N) can be neglected relative to order 1.
The entire analytical theory, Eqs. 4-6, is derived in this limit (Section III); agreement with numerics at high affinity is the evidence for this assumption.
ad hoc to paper Matching the mean and variance of the first-passage-time distribution across a decoration to a single link preserves the global period and coherence of oscillations.
Section IV states the expectation that T is 'roughly captured by the first moment' and R by the second; no proof that higher moments are irrelevant is given.
domain assumption Decorations are small (at most 6 sides), non-overlapping, and in either cis or trans rate configurations, so effective coarse-grained rates remain positive and decorations do not support independent oscillations.
Section IV restricts the study to this regime; the effective rates diverge or become negative outside it (Table I).
ad hoc to paper The perturbation replacement X(0)_1 -> X(1)_1 in Appendix A is valid when coarse-grained links have h-/h+ > 1.
Appendix A introduces this replacement saying 'we find' that previously ignored gamma terms cancel, but the cancellation is not shown; the analytical predictions for cis decorations depend on it.
standard math Laplace-transformed first-passage time distributions are obtained by matrix inversion of (1-K)^-1 and moments by derivatives at s=0.
Appendix B uses standard Markov chain first-passage time theory; this is a textbook result.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Robust oscillations in multi-cyclic models of biochemical clocks." pith.science (2026). https://pith.science/paper/EXAYGK4Y

@misc{pith2026190902534,
  author       = {Pith},
  title        = {Pith review of: Robust oscillations in multi-cyclic models of biochemical clocks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/EXAYGK4Y}},
  note         = {Machine review of arXiv:1909.02534}
}

read the original abstract

Organisms often use cyclic changes in the concentrations of chemicals species to precisely time biological functions. Underlying these biochemical clocks are chemical reactions and transport processes, which are inherently stochastic. Understanding the physical basis for robust biochemical oscillations in the presence of fluctuations has thus emerged as an important problem. In a previous paper [C. del Junco and S. Vaikuntanathan, Phys. Rev. E 101, 012410 (2020)], we explored this question using the non-equilibrium statistical mechanics of single-ring Markov state models of biochemical networks that support oscillations. Our finding was that they can exploit non-equilibrium driving to robustly maintain the period and coherence of oscillations in the presence of randomness in the rates. Here, we extend our work to Markov state models consisting of a large cycle decorated with multiple small cycles. These additional cycles are intended to represent alternate pathways that the oscillator may take as it fluctuates about its average path. Combining a mapping to single-cycle networks based on first passage time distributions with our previously developed theory, we are able to make analytical predictions for the period and coherence of oscillations in these networks. One implication of our predictions is that a high energy budget can make different network topologies and arrangements of rates degenerate as far as the period and coherence of oscillations is concerned. Excellent agreement between analytical and numerical results confirms that this is the case. Our results suggest that biochemical oscillators can be more robust to fluctuations in the path of the oscillator when they have a high energy budget.

Figures

Figures reproduced from arXiv: 1909.02534 by the authors.

**Figure 2.** FIG. 2. Effective rates resulting from the coarse-graining pro [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

**Figure 4.** FIG. 4. Period [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗

Figures from the paper (1 more)

**Figure 6.** Figure 6: FIG. 6. Compensating for changes in period by tuning the pa [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

22 extracted references · 21 canonical work pages

[1]

M. A. Woelﬂe, Y. Ouyang, K. Phanvijhitsiri, and C. H. Johnson, Curr. Biol. 14, 1481 (2004)

work page 2004
[2]

Nakajima, K

M. Nakajima, K. Imai, H. Ito, T. Nishiwaki, Y. Mu- rayama, H. Iwasaki, T. Oyama, and T. Kondo, Science 308, 414 (2005)

work page 2005
[3]

Tomita, M

J. Tomita, M. Nakajima, T. Kondo, and H. Iwasaki, Science 307, 251 (2005)

work page 2005
[4]

M. J. Rust, J. S. Markson, W. S. Lane, D. S. Fisher, and E. K. O’Shea, Science 318, 809 (2007)

work page 2007
[5]

A. C. Barato and U. Seifert, Phys. Rev. Lett.114, 158101 (2015)

2015
[6]

Y. Cao, H. Wang, Q. Ouyang, and Y. Tu, Nat. Phys. 11, 772 (2015)

work page 2015
[7]

A. C. Barato and U. Seifert, Phys. Rev. E 95, 062409 (2017)

work page 2017
[8]

C. Fei, Y. Cao, Q. Ouyang, and Y. Tu, Nat. Commun. 9, 1434 (2018)

work page 2018

Show all 22 references

[9]

Wierenga, P

H. Wierenga, P. R. Ten Wolde, and N. B. Becker, Phys. Rev. E 97, 042404 (2018)

2018
[10]

Nguyen, U

B. Nguyen, U. Seifert, and A. C. Barato, J. Chem. Phys. 149, 045101 (2018)

2018
[11]

Marsland, W

R. Marsland, W. Cui, and J. M. Horowitz, J. R. Soc. Interface 16, 20190098 (2019)

2019
[12]

del Junco and S

C. del Junco and S. Vaikuntanathan, Phys. Rev. E 101, 012410 (2020)

2020
[13]

Nov´ ak and J

B. Nov´ ak and J. J. Tyson, Nat. Rev. Mol. Cell Biol. 9, 981 (2008)

2008
[14]

12 and 7, φ was deﬁned as the eigenvalue with the least negative real part

In refs. 12 and 7, φ was deﬁned as the eigenvalue with the least negative real part. However, in the multi-cyclic networks we address later in this paper, that deﬁnition can lead to selecting an eigenvalue which corresponds to cycling around a decoration rather than global osc...
[15]

Murugan, D

A. Murugan, D. A. Huse, and S. Leibler, Proc. Natl. Acad. Sci. U. S. A. 109, 12034 (2012)

2012
[16]

Budnar, K

S. Budnar, K. B. Husain, G. A. Gomez, M. Naghibosa- dat, A. Varma, S. Verma, N. A. Hamilton, R. G. Morris, and A. S. Yap, Dev. Cell 49, 894 (2019)

2019
[17]

Pittayakanchit, Z

W. Pittayakanchit, Z. Lu, J. Chew, M. J. Rust, and A. Murugan, Elife 7 (2018), 10.7554/eLife.37624

2018 doi
[18]

C. H. Johnson, P. L. Stewart, and M. Egli, Annu. Rev. Biophys. 40, 143 (2011)

2011
[19]

Fran¸ cois, N

P. Fran¸ cois, N. Despierre, and E. D. Siggia, PLoS Com- put. Biol. 8, e1002585 (2012)

2012
[20]

Paijmans, D

J. Paijmans, D. K. Lubensky, and P. R. ten Wolde, Bio- phys. J. 113, 157 (2017)

2017
[21]

T. S. Hatakeyama and K. Kaneko, Phys. Rev. Lett. 115, 218101 (2015). Appendix A: Theory from Ref. 12 Our theory in Ref. 12 uses a transfer matrix formu- lation of the eigenvalue equation for the transition rate matrix of a single-cycle network of size N where all but one set o...

2015
[22]

The ﬁrst term is the probability of hopping at exactly time t1, while the term in parentheses is the probability that the system has not hopped to any other state in the meantime

If the system enters state 1 at time t0, the probability that it hops to state j at time t1 =t0 +µt is: Q12(µt) =P12(µt) ∏ i⁄=1,2 ( 1− ∫ µt 0 dtP1i(t) ) (B1) Where P12(t) = k12 exp(−k12t) is the waiting time dis- tribution for hopping from state 1 to 2. The ﬁrst term is the pr...

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [1]

M. A. Woelﬂe, Y. Ouyang, K. Phanvijhitsiri, and C. H. Johnson, Curr. Biol. 14, 1481 (2004)

work page 2004

[2] [2]

Nakajima, K

M. Nakajima, K. Imai, H. Ito, T. Nishiwaki, Y. Mu- rayama, H. Iwasaki, T. Oyama, and T. Kondo, Science 308, 414 (2005)

work page 2005

[3] [3]

Tomita, M

J. Tomita, M. Nakajima, T. Kondo, and H. Iwasaki, Science 307, 251 (2005)

work page 2005

[4] [4]

M. J. Rust, J. S. Markson, W. S. Lane, D. S. Fisher, and E. K. O’Shea, Science 318, 809 (2007)

work page 2007

[5] [5]

A. C. Barato and U. Seifert, Phys. Rev. Lett.114, 158101 (2015)

2015

[6] [6]

Y. Cao, H. Wang, Q. Ouyang, and Y. Tu, Nat. Phys. 11, 772 (2015)

work page 2015

[7] [7]

A. C. Barato and U. Seifert, Phys. Rev. E 95, 062409 (2017)

work page 2017

[8] [8]

C. Fei, Y. Cao, Q. Ouyang, and Y. Tu, Nat. Commun. 9, 1434 (2018)

work page 2018