Pith. sign in

REVIEW 3 major objections 3 minor 39 references

The paper's central claim is that a single receptor's information capacity is unbounded under ideal timing, but becomes finite once the timing precision of its averaging clock is taken into account.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-01 08:02 UTC pith:NA4LTJ36

load-bearing objection Clean closed-form link between Berg-Purcell precision and information capacity, with a real clock-limited saturation result, but the 'bound' wording outruns the math: the central formula is an asymptotic offset and an underestimate, not an upper bound. the 3 major comments →

arxiv 2607.21249 v1 pith:NA4LTJ36 submitted 2026-07-23 q-bio.MN cs.ITmath.ITstat.AP

From Berg-Purcell precision bounds to clock-limited information capacity

Micha{\l} Komorowski This is my paper
classification q-bio.MN cs.ITmath.ITstat.AP
keywords information capacitytwo-state receptorchemical sensingstochastic clockdynamic rangetiming precisionoccupancy averagingmutual information
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper asks how many bits, not just how accurately, a single molecular receptor can report a chemical concentration. It derives a compact formula that splits the receptor's information capacity into a dynamic-range term, an averaging-time term, and a copy-number term. In the idealized case where the integration window is known exactly, the formula shows capacity grows without bound as the allowed concentration range widens, but only roughly as the log of the log of range, so additional bits are extremely expensive. Once the receptor's internal clock has finite timing precision, high-concentration discrimination degrades faster and the total capacity saturates at a finite number of bits. This matters because experiments increasingly measure signaling in bits, and it explains why low single-bit capacities do not require a failure of receptor-level physics.

Core claim

The central claim is that local concentration-precision limits do not, by themselves, set a finite global information budget for a receptor. Under ideal fixed-time averaging, the information capacity grows without bound with the upper end of the input range, scaling as log2(ln(c_max/K_d)). The paper then treats the duration used to normalize the occupancy readout as a stochastic clock with known mean and variance. Clock noise changes the high-concentration scaling of the discrimination precision from c^-2 to c^-4, which makes the capacity integral converge. The resulting closed-form bound, expressed through incomplete elliptic integrals, is finite even for an unbounded concentration range an

What carries the argument

The carrying object is the asymptotic capacity formula C*_A = log2( (1/sqrt(2πe)) ∫_{cmin}^{cmax} sqrt(I(c)) dc ), which converts a local measure of how sharply the readout distinguishes nearby concentrations (I(c)) into a global, distribution-free capacity in bits. Applied to time-averaged receptor occupancy, it yields Eq. (8) for ideal timing, separating range, time, and copy-number contributions. To make the ideal model physical, the paper introduces a noisy clock: the downstream circuit normalizes the accumulated bound time by an imperfect estimate with mean T and variance σ_T^2. This adds a variance term that changes I(c) at high occupancy and makes the integral evaluable as an incomple

Load-bearing premise

The whole calculation rests on a standard large-sample formula that converts local discrimination precision into information capacity; if that approximation is not accurate for a single receptor observed over a finite window, the quoted bit values are asymptotic offsets rather than true information bounds.

What would settle it

Simulate the two-state receptor with a stochastic Erlang clock, compute the exact mutual information between concentration and the occupancy readout for a single receptor over a bounded range, and check whether the large-sample extrapolation matches Eq. (13); if capacity still grows substantially with cmax, or if the plateau height does not follow the 1/sqrt(m) timing-precision scaling, the clock-limited bound is falsified.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Doubling the averaging time or the receptor copy number adds only half a bit; a full additional bit requires a fourfold increase.
  • Widening the dynamic range gives only log-logarithmic gains in the ideal model, so an unbounded range never yields a practical route to many bits.
  • With finite clock precision, the capacity plateaus in both averaging time and dynamic range, so the limiting resource becomes timing precision rather than diffusive sampling.
  • The predicted low single-bit capacities are consistent with the range seen in biological signaling measurements, suggesting receptors can operate near their physical limit while downstream information remains low.

Where Pith is reading between the lines

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

  • I would extend the logic beyond biochemistry: any physical sensor that counts stochastic events and normalizes by a measured duration should exhibit the same crossover, so Eq. (13) may serve as a generic bound for counting sensors with noisy clocks.
  • A consequence the author leaves implicit: comparing 'information capacities' of different receptors is only meaningful when dynamic range and timing precision are specified, because otherwise the ideal-model capacity is unbounded.
  • Testable extension: if the clock is an m-step Erlang timer, the timing-precision coefficient equals 1/sqrt(m); one could engineer timers with different m in synthetic circuits and check that the capacity plateau shifts as predicted.
  • The model assumes clock noise is independent of binding state; coupling between the two, such as a timer driven by shared biochemical resources, could create correlations that either raise or lower the bound.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 3 minor

Summary. The paper derives a Fisher-information-based asymptotic capacity formula for a two-state receptor. In the ideal fixed-time occupancy model, it obtains a closed-form capacity expression showing slow (log-log) unbounded growth with concentration dynamic range. The paper then introduces a noisy internal clock that perturbs the normalization time, derives a modified Fisher information with an additional variance term, evaluates the capacity integral in closed form using incomplete elliptic integrals, and claims that finite timing precision yields a finite, clock-limited receptor-level information bound. The central conceptual claim is that diffusion-limited sampling alone does not produce a finite global information bound unless timing precision or some other physical resource is included.

Significance. If the asymptotic capacity interpretation is accepted, the paper provides useful, analytically transparent formulas that separate dynamic-range, averaging-time, and copy-number contributions, and it identifies a robust qualitative crossover from diffusion-limited (c^-2) to clock-limited (c^-4) scaling of the Fisher information at high occupancy. The supplementary derivations are internally consistent, the elliptic-integral evaluation is correct, and the log-log divergence of the ideal fixed-time capacity is a clean observation. However, the central 'bound' status of Eq. (13) is not established by the derivation as written, so the reported bit values and the upper-bound language overstate what has been proven.

major comments (3)
  1. [Main text, Eq. (13); Supplement S1.1] Eq. (13) is presented as a 'receptor-level information bound in bits,' but it is computed from the Clarke–Barron asymptotic capacity offset of Eq. (4), which is an N→∞ offset, not an upper bound on the finite-N Shannon capacity. The figures plot C_A^*, not C_N^*, for N=1, and the authors themselves note that C_A^* can be negative. Thus Eq. (13) does not, by itself, bound the mutual information of a single receptor. The authors need either to prove a genuine finite-N upper bound or to explicitly recast the claims and figures as statements about the asymptotic offset rather than about bits available to a single receptor.
  2. [Supplement S4.3, Eqs. (S30)–(S31); Main text Eq. (10)] The Fisher information is obtained by inverting a delta-method variance, I_eff(c)=1/Var(ĉ). For a Gaussian observation with a concentration-dependent variance v(c), the exact Fisher information is (p')^2/v + (v')^2/(2v^2) ≥ (p')^2/v, so I_eff(c) ≤ I_true(c) pointwise. Consequently ∫√I_eff dc ≤ ∫√I_true dc, and Eq. (13) is a lower estimate of the occupancy-channel capacity, not an upper bound. The statement that 'clock-imprecise capacity is always below the ideal-timing bound' compares two effective-FI quantities and does not establish an upper bound on the exact capacity. The c^-4 saturation mechanism is likely robust, but the bit values and the word 'bound' are not supported by the derivation as written.
  3. [Supplement S1.1 and S2; Main text Eqs. (7)–(8)] The asymptotic capacity formula (S4) is derived for M i.i.d. samples via the Bernstein–von Mises theorem, but here the observation is a single time-averaged occupancy trajectory, and the Fisher information in Eq. (7) is itself obtained from a long-time variance approximation. No error estimate is given for combining these two asymptotic approximations. At finite T, Eq. (13) therefore has unknown accuracy, even as an approximation of the asymptotic capacity. The authors should either supply a more direct justification for applying Eq. (S4) to this continuous-time single-trajectory setting or explicitly state the double-asymptotic nature of the result.
minor comments (3)
  1. [Eq. (12)] The elliptic-integral parameter μ=1−s^2 k_- T/8 can become negative when s^2 k_- T>8. Specify the parameter convention for F(φ|μ) used here, since many readers expect the incomplete elliptic integral with parameter in [0,1].
  2. [Figures 2 and 3] The axis labels and captions plot C_A^*, the asymptotic offset, for N=1. Consider relabeling the axes as 'asymptotic capacity offset' rather than 'capacity' to avoid any impression that negative values are actual negative mutual information, and to make the N=1 caveat more prominent.
  3. [Supplement S4.2] The derivation linearizes the clock error in δT_v/T. For the Erlang clock with small m (e.g., m=10), the coefficient of variation s=1/√m is not very small. It would be helpful to state explicitly that the result is a small-noise approximation and possibly to test it against a numerical evaluation of the exact Erlang-clock mixture.

Circularity Check

0 steps flagged

No significant circularity: the FI-to-capacity bridge is independently derived in the supplement, and the noisy-clock regularization is an explicit model assumption rather than a disguised fit.

full rationale

The derivation chain is: FI for the two-state receptor from BP variance (Eq. 7 / S14), conversion of FI to asymptotic capacity via Eq. 4, integration to Eq. 8, and stochastic-clock correction leading to Eq. 13. None of these steps reduces to its own target by construction. Eq. (4) is cited to Refs. [17,18], but the supplement (S1.1) re-derives it from the Bernstein–von Mises expansion of mutual information, and the formula is the standard Clarke–Barron/Jeffreys asymptotic capacity; the self-citations are therefore not load-bearing. The clock model is introduced as an explicit stated assumption ('We therefore introduce finite timing precision as a physically motivated completion...'), not fitted to the output it predicts, and the derivation of I_clock (S4.3) follows from variance propagation rather than from the capacity formula. The saturation behavior arises algebraically from the assumed c^-4 FI scaling and is not equivalent to the ideal-timing result. Whether Eq. (13) is a true finite-N upper bound is an asymptotic/approximation concern, not a circularity concern.

Axiom & Free-Parameter Ledger

2 free parameters · 4 axioms · 0 invented entities

The paper fits no data; all calculations are analytic. The key imported result is the Fisher-information-to-capacity formula from refs. [17,18], and the key new modeling choice is the stochastic clock. The clock model is introduced as a regularization, so the finite bound is conditional on that model.

free parameters (2)
  • s^2 = (sigma_T/T)^2 (relative clock variance) = symbolic; Erlang clock gives s = 1/sqrt(m)
    Chosen by hand to model timing precision; the saturation result requires s > 0, and the value of the clock-limited bound depends on it. Not fitted to data.
  • m (Erlang timer steps) = free integer; examples 10, 100, 1000 in figures
    Controls clock precision via s = 1/sqrt(m); represents biochemical resources invested in temporal precision but no empirical estimate is given.
axioms (4)
  • domain assumption Asymptotic capacity formula C*_A = log2((1/sqrt(2*pi*e)) * integral sqrt(I(c)) dc) from refs. [17,18] is a valid approximation of Shannon capacity in the large-N regime.
    Every capacity statement in the paper is computed through this equation; it is cited from prior work (two refs by the same author) rather than proved here, and it can be negative, so it is an asymptotic offset rather than an exact bound.
  • domain assumption Time-averaged occupancy Y is approximately Gaussian with Var(Y) ~ 2 p(1-p) tau_c / T for large T, and efficient estimation reaches the Cramér–Rao bound so FI ~ 1/Var(c_hat).
    Used in Supplementary Sec. S2 to obtain Fisher information (Eq. 7); standard in Berg–Purcell theory but an approximation that neglects higher-order terms and non-Gaussianity.
  • ad hoc to paper The noisy clock T_v is independent of receptor switching, has mean T and variance sigma_T^2, and enters only through a linearized denominator correction; sensing and clock variances add.
    Introduced in Supplementary Secs. S4.1–S4.2; this specific regularization produces the c^-4 Fisher-information scaling and the saturation. Different timing-noise models would give different bounds.
  • domain assumption The receptor is a stationary two-state Markov process with equilibrium occupancy p(c) = c/(c+K_d).
    Canonical model from refs. [1–4]; underlies all Fisher-information calculations in the paper.

pith-pipeline@v1.3.0-alltime-deepseek · 9752 in / 23322 out tokens · 245463 ms · 2026-08-01T08:02:21.104425+00:00 · methodology

0 comments
read the original abstract

Physical limits to chemical sensing are traditionally expressed as Berg-Purcell bounds on estimation accuracy. Whether these bounds also limit the total amount of information a molecular receptor can transmit has remained unclear, despite the fact that cellular signaling performance is naturally quantified in bits rather than precision alone. Here we derive an explicit link between Berg-Purcell-type sensing limits and the information capacity of a single two-state receptor, yielding a compact expression that separates contributions from concentration range, receptor copy number, and averaging time. We show that, in the ideal fixed-time occupancy model, diffusion-limited sampling alone does not define a finite global information bound: with a perfectly specified integration window and unbounded input range, information capacity grows without bound with dynamic range, albeit slowly. A finite saturation arises when time integration is treated as an explicit physical resource. Finite timing precision yields a clock-limited bound on information capacity, and in the high-occupancy regime information transmission crosses over from diffusion-limited to clock-limited behavior. Together, these results establish a receptor-level information bound in bits that is finite once constraints on timing precision are taken into account.

Figures

Figures reproduced from arXiv: 2607.21249 by Micha{\l} Komorowski.

Figure 1
Figure 1. Figure 1: FIG. 1. Single-receptor trajectory and the physical origin [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. Capacity ( [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

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

39 extracted references

  1. [1]

    H. C. Berg and E. M. Purcell, Biophysical journal20, 193 (1977)

    1977

  2. [2]

    Bialek and S

    W. Bialek and S. Setayeshgar, Proceedings of the Na- tional Academy of Sciences102, 10040 (2005)

    2005

  3. [3]

    Mora, Physical review letters115, 038102 (2015)

    T. Mora, Physical review letters115, 038102 (2015)

    2015

  4. [4]

    P. R. ten Wolde, N. B. Becker, T. E. Ouldridge, and A. Mugler, Journal of Statistical Physics162, 1395 FIG. 1. Single-receptor trajectory and the physical origin of a clock-limited information bound. Idealized Berg–Purcell sensing assumes the cell forms the time-averaged occupancy Y=T B/Tover a well-defined integration windowT. This schematic highlights t...

    2016

  5. [5]

    R. G. Endres and N. S. Wingreen, Physical review letters 103, 158101 (2009)

    2009

  6. [6]

    Kaizu, W

    K. Kaizu, W. De Ronde, J. Paijmans, K. Takahashi, F. Tostevin, and P. R. Ten Wolde, Biophysical journal 106, 976 (2014)

    2014

  7. [7]

    Aquino, N

    G. Aquino, N. S. Wingreen, and R. G. Endres, Journal of statistical physics162, 1353 (2016)

    2016

  8. [8]

    Handy and S

    G. Handy and S. D. Lawley, Biophysical journal120, 2237 (2021)

    2021

  9. [9]

    C. E. Shannon, The Bell system technical journal27, 379 (1948)

    1948

  10. [10]

    T. M. Cover,Elements of information theory(John Wi- ley & Sons, 1999)

    1999

  11. [11]

    Cheong, A

    R. Cheong, A. Rhee, C. J. Wang, I. Nemenman, and A. Levchenko, science334, 354 (2011)

    2011

  12. [12]

    Selimkhanov, B

    J. Selimkhanov, B. Taylor, J. Yao, A. Pilko, J. Albeck, A. Hoffmann, L. Tsimring, and R. Wollman, Science346, 1370 (2014)

    2014

  13. [13]

    Tkaˇ cik and P

    G. Tkaˇ cik and P. R. Ten Wolde, Annual review of bio- physics54, 249 (2025)

    2025

  14. [14]

    Jeffreys, Proceedings of the Royal Society of London

    H. Jeffreys, Proceedings of the Royal Society of London. Series A. Mathematical and Physical Sciences186, 453 (1946)

    1946

  15. [15]

    B. S. Clarke and A. R. Barron, IEEE Transactions on information theory36, 453 (2002)

    2002

  16. [16]

    Brunel and J.-P

    N. Brunel and J.-P. Nadal, Neural computation10, 1731 (1998)

    1998

  17. [17]

    Jetka, K

    T. Jetka, K. Niena ltowski, S. Filippi, M. P. Stumpf, and M. Komorowski, Nature communications9, 4591 (2018)

    2018

  18. [18]

    Komorowski, Communications Physics (2025)

    M. Komorowski, Communications Physics (2025). 5 FIG. 2. Capacity (C ∗ A) vs mean integration timeTfor a fixed dynamic range. HereC ∗ A is the additive asymptotic offset in Eq. (5), plotted for a single receptor (N= 1), so the fullN-receptor capacity isC ∗ N =C ∗ A + 1 2 log2 N; neg- ative values indicate the effectively zero-bit single-receptor regime, no...

    2025

  19. [19]

    R. A. Fisher, Mathematical proceedings of the Cam- bridge philosophical society22, 700 (1925)

    1925

  20. [20]

    C. R. Raoet al., Bull. Calcutta Math. Soc37, 81 (1945)

    1945

  21. [21]

    Cram´ er,Mathematical methods of statistics, Vol

    H. Cram´ er,Mathematical methods of statistics, Vol. 9 (Princeton university press, 1999)

    1999

  22. [22]

    Le Cam,Asymptotic methods in statistical decision theory(Springer Science & Business Media, 2012)

    L. Le Cam,Asymptotic methods in statistical decision theory(Springer Science & Business Media, 2012)

    2012

  23. [23]

    Fancher and A

    S. Fancher and A. Mugler, Physical review letters118, 078101 (2017)

    2017

  24. [24]

    Tostevin and P

    F. Tostevin and P. R. ten Wolde, Physical Review Letters 102, 218101 (2009)

    2009

  25. [25]

    Mora and I

    T. Mora and I. Nemenman, Physical review letters123, 198101 (2019)

    2019

  26. [26]

    Tkaˇ cik, C

    G. Tkaˇ cik, C. G. Callan Jr, and W. Bialek, Proceedings of the National Academy of Sciences105, 12265 (2008)

    2008

  27. [27]

    Tkaˇ cik, J

    G. Tkaˇ cik, J. O. Dubuis, M. D. Petkova, and T. Gregor, Genetics199, 39 (2015)

    2015

  28. [28]

    Celani and M

    A. Celani and M. Vergassola, Proceedings of the National Academy of Sciences107, 1391 (2010)

    2010

  29. [29]

    D. R. Brumley, F. Carrara, A. M. Hein, Y. Yawata, S. A. Levin, and R. Stocker, Proceedings of the National Academy of Sciences116, 10792 (2019)

    2019

  30. [30]

    Micali, R

    G. Micali, R. Colin, V. Sourjik, and R. G. Endres, Bio- physical journal113, 2321 (2017)

    2017

  31. [31]

    K. Y. Wan and G. J´ ekely, Philosophical Transactions of the Royal Society B376, 20190758 (2021)

    2021

  32. [32]

    H. H. Mattingly, K. Kamino, J. Ong, R. Kottou, T. Emonet, and B. B. Machta, Nature Physics , 1 (2026)

    2026

  33. [33]

    Tu and G

    Y. Tu and G. Grinstein, Physical review letters94, 208101 (2005). FIG. 3. Capacity (C ∗ A) vs dynamic rangec max/cmin at fixed T. With ideal timing (solid), Berg–Purcell scaling predicts a slow but unbounded growth of capacity,C ∗ A ∼log 2 lnc max (Eq. (8)). Finite timing precision (dashed/dotted) changes the high-concentration scaling of the Fisher infor...

    2005

  34. [34]

    Matth¨ aus, M

    F. Matth¨ aus, M. S. Mommer, T. Curk, and J. Dobnikar, PloS one6, e18623 (2011)

    2011

  35. [35]

    Topolewski and M

    P. Topolewski and M. Komorowski, Current Opinion in Systems Biology27, 100352 (2021)

    2021

  36. [36]

    Komorowski and D

    M. Komorowski and D. S. Tawfik, Cell systems8, 76 (2019)

    2019

  37. [37]

    Carballo-Pacheco, J

    M. Carballo-Pacheco, J. Desponds, T. Gavrilchenko, A. Mayer, R. Prizak, G. Reddy, I. Nemenman, and T. Mora, Physical Review E99, 022423 (2019)

    2019

  38. [38]

    Y. E. Antebi, J. M. Linton, H. Klumpe, B. Bintu, M. Gong, C. Su, R. McCardell, and M. B. Elowitz, Cell 170, 1184 (2017)

    2017

  39. [39]

    effectively zero bits

    I. Amit, R. Wides, and Y. Yarden, Molecular systems biology3, 151 (2007). ACKNOWLEDGEMENTS This research was supported by the National Science Center, Poland, under grant number 2020/39/B/NZ2/03259. AUTHOR CONTRIBUTIONS M.K. conceived and designed the study, performed the analyses, and wrote the manuscript. COMPETING INTERESTS The author declares no compe...

    2007