{"id":"4f745d3c-0e42-474c-854e-9e9c21cab76e","arxiv_id":"2412.21113","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"In a one-dimensional exclusion process with position-dependent Langmuir kinetics, optimal trade-offs between positional information, entropy production, and reaction current exhibit kinks, phase transitions, and diminishing returns.","lead":"This paper derives trade-offs between how much positional information a boundary-driven particle system can carry, how much energy it must dissipate, and how fast reactions occur. It finds that optimal protocols switch discontinuously between different boundary conditions, like phase transitions, and that adding a reaction cluster can boost the maximum positional information.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The dark-region first-order transition and 'thermodynamically suboptimal zone' rest on a concave Pareto segment that linear scalarization (Eq. 34) cannot resolve; the segment is inferred from optimizer jumps and missing SM, not computed.","rationale":"The exact derivations for the clustered and uniform profiles are largely self-contained: the density profile, the positional-information formula, and the exact Sigma-J trade-offs follow from Eqs. (7), (18), (23), (31)-(33), so the existence of non-trivial trade-offs is on solid ground. The fragile part is the claim that the dark-region PI-Sigma front contains a concave segment and a first-order transition. The reader identified this as the weakest assumption, and my reading agrees: Eq. (34) is linear scalarization, a support-function method that locates the convex hull of the feasible set but cannot by itself certify or trace a concave part of the Pareto front. The paper's evidence for that concave part consists of the observed jump in the optimizer, 'metastable' profiles, and a SM that is missing from the submitted source ('??' placeholders). This is an inference, not a computation. The proposed epsilon-constraint test settles it by computing the actual lower envelope. I do not see a separate fundamental flaw in the hydrodynamic derivation or the other exact trade-offs, so the conditional verdict remains appropriate.","tokens_in":17775,"tokens_out":13644,"duration_ms":146876,"concrete_test":"Recompute the dark-region Pareto front directly by epsilon-constraint optimization: for gamma = 0.1, alpha = 9 (the cyan point in Fig. 5), choose a dense set of target information values I* spanning the interval between the two branch endpoints at lambda_c,2, and for each I* minimize Sigma over (rho_L, rho_R) in [0,1]^2 subject to I(rho_L, rho_R) >= I*, using a fine grid or a global optimizer. Plot the resulting lower envelope together with the lambda-sweep of Eq. (34). If the epsilon-constrained minima contain non-dominated points that no lambda selects (a non-convex arc away from the scalarization path), the concave-segment claim is confirmed; if the lower envelope is convex and coincides with the scalarization curve, the first-order transition and suboptimal-zone interpretation should be revised.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central novel claim most at risk is the first-order 'phase transition' and the 'thermodynamically suboptimal zone' in the uniform-profile dark region (Sec. IV B, Figs. 7(b), 8). The authors obtain this from the scalarized objective Omega = -lambda I + (1-lambda) Sigma, Eq. (34), sweeping lambda from 0 to 1. For a two-objective problem, minimizing a positive linear combination only returns supported points on the convex hull of the feasible set; it cannot return points lying on a concave (non-convex) segment of the true Pareto front. In the dark region, the concave segment near lambda_c,2 = 0.967 is inferred from the optimizer jumping from monotonic to non-monotonic density profiles and from claimed 'metastable' profiles, rather than from a direct computation of the lower envelope. The exact expression for the cusp is relegated to Eq. (??) of a Supplemental Material that is absent from the submitted source, so the key step is not independently checkable. If that inferred segment is not actually a set of non-dominated (rho_L, rho_R) profiles, then the exchange of global optimality and the 'suboptimal zone' do not exist as stated, and the phase diagram in Fig. 8 loses its first-order branch.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies a one-dimensional boundary-driven symmetric simple exclusion process (SSEP) with position-dependent Langmuir kinetics, in the hydrodynamic limit. It derives exact steady-state density profiles for clustered (delta-function) and uniform enzyme profiles, and uses them to compute the positional information I (mutual information between site index and occupancy), the rescaled entropy production rate Σ, and the global reaction current J. The authors then analyze Pareto-optimal trade-offs among these quantities when the boundary reservoir densities ρL, ρR are tuned. For the clustered profile, they report a maximum positional information Im = log2((1+e)/e) ≈ 0.452 bits, exceeding the previous no-Langmuir bound of 0.278 bits. For the uniform profile, they identify two regimes in the (γ, α) parameter space: a 'light' regime with a convex I–Σ Pareto front and a second-order transition in the optimal protocol, and a 'dark' regime with a concave segment of the front, a first-order transition at λc,2, and a claimed 'thermodynamically suboptimal zone.' They also present a WKB approximation for general slowly varying enzyme profiles. The central claims are that non-trivial Pareto-optimal trade-offs exist among I, Σ, and J, and that optimal protocols exhibit phase transitions analogous to liquid-gas coexistence.","tokens_in":18036,"tokens_out":4206,"duration_ms":43311,"significance":"If the results hold, this paper provides a concrete, exactly solvable statistical-mechanics model connecting positional information, dissipation, and reaction flux, extending earlier work on information-thermodynamics trade-offs in boundary-driven systems. The exact closed-form expressions for the Σ–J Pareto fronts (Eqs. 33 and 41) and the explicit maximum for the clustered profile are valuable, as is the WKB treatment for general profiles. The model is cleanly defined, and the derivations of the density equation and the observables are transparent and do not rely on free parameters or fitted data. The paper also makes contact with recent ideas on thermodynamic costs of communication and on inverse multiplexing. However, the central phase-transition claim in the dark region of the uniform-profile case currently rests on a concave Pareto segment that is inferred rather than directly computed, and several load-bearing derivations are relegated to a Supplemental Material that is missing from the submitted source. These issues are significant but appear fixable, so the paper is a candidate for major revision rather than rejection.","major_comments":[{"comment":"The first-order phase transition and the 'thermodynamically suboptimal zone' in the dark region are not established by the method used. The authors minimize the scalarized objective Ω = −λI + (1−λ)Σ (Eq. 34) and sweep λ. Linear scalarization only returns supported points on the convex hull of the feasible objective set; it cannot produce points on a concave (non-convex) segment of the true Pareto front. The paper itself states that at λc,2 the front 'becomes locally concave' and that the transition is inferred from the optimizer jumping from monotonic to non-monotonic profiles plus 'metastable' profiles, rather than from a direct computation of the lower envelope. The exact expression for the cusp is deferred to Eq. (??) of the SM, which is absent. As written, the existence of the first-order branch in Fig. 8 and the suboptimal-zone claim are unsupported. To make this load-bearing claim, the authors should compute the Pareto front directly in the concave region (e.g., by minimizing Σ for fixed I, or by a normal-boundary/ε-constraint method) and show that the inferred segment is indeed non-dominated.","section":"§IV A, Eq. (30)"},{"comment":"The headline result for the clustered profile, Im = log2((1+e)/e) ≈ 0.452 bits, is stated without a derivation in the main text; the derivation is said to be in 'section ?? of the SM,' which is missing from the submitted source. The heuristic argument that follows (ρL=1, ρR=0, γ→0, α→∞, κ0=1−tanh(1/2)) is plausible, but the limit involves a singular perturbation of Eq. (7) with a delta-function source, and the order of limits matters. The result should be either derived in the main text or provided in a complete, accessible SM. This is a central quantitative claim of the paper and must be independently checkable.","section":"§II, Eq. (7) through §IV B"},{"comment":"The manuscript is not self-contained: multiple key expressions are referenced as 'eq. (??)' or 'section ?? of the SM [51]' (e.g., the exact PI for the clustered profile, the derivation of Im, the cusp solution in the dark region, and the boundary-derivative formulas for general profiles). The Supplemental Material is not present in the source. Since several of these omitted items are load-bearing for the main claims, this is a major completeness issue, not merely a typographical one. The authors must supply the SM and fill in all placeholder references before the paper can be evaluated.","section":"§IV C, Eq. (44) and Fig. 9"},{"comment":"The WKB approximation is advertised as yielding 'qualitatively accurate trade-offs for general Langmuir density profiles,' but the only quantitative comparison is the fractional change of the maximal PI (inset of Fig. 9) for a linear profile, for three values of α. It is not shown that the full Pareto front (not just the maximum) is reproduced within the stated O(α^{-1}) error. This is a lesser point, but the comparison should be extended to the whole front or the statement should be restricted to the maximum PI.","section":"§IV A, Eq. (30)"}],"minor_comments":[{"comment":"The sentence 'any solution that minimises eq. (34) is on the Pareto optimal front' is only true for supported Pareto points; for non-convex fronts it is false. Since the paper later discusses a concave front, this statement should be qualified accordingly.","section":"§IV B, around Fig. 7(b)"},{"comment":"The term 'metastable' is applied to optimal density profiles that are not global minima of Ω. In this steady-state optimization, hysteresis in the optimizer as λ is swept does not by itself imply physical metastability; the terminology should be justified or replaced with 'locally optimal' or 'suboptimal branches.'","section":"§II, before Eq. (7)"},{"comment":"The rescaling of Langmuir rates (kon = ωon/N^2, koff = ωoff/N^2) is introduced in prose, but Eq. (7) then uses α^2 = (ωon+ωoff)/p without recalling the definition; restating it directly after Eq. (7) would improve readability.","section":"§III, Eq. (25)"},{"comment":"In Eq. (25), the same symbol J is used for the rescaled and unrescaled reaction flux; the text says 'J = J N/(ωon+ωoff)' but the equation then writes J on both sides. This notation collision should be cleaned up.","section":"§IV C"},{"comment":"The WKB solution (44) uses the notation csch(F(0,1)) and F(s,t) = α∫_s^t dq sqrt(e(q)); ensure that the argument of csch is dimensionless and that the approximation is stated for e(κ) strictly positive, since the square root requires it.","section":"Conclusion and Data availability"},{"comment":"The data availability statement says all data are in the article and any supplementary files, but the SM is currently missing; this statement should be updated once the SM is provided.","section":"Throughout"},{"comment":"There are numerous '??' placeholders for equation numbers and SM section references (e.g., after Eqs. (29), (32), (39), (44)). These must be resolved before publication.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The paper presents a clean hydrodynamic model and several exact results that are likely of interest to the statistical-physics community. The main risk is the unsupported concave Pareto segment in the dark region; if the authors can provide a direct construction of that segment (or appropriately weaken the phase-transition claim), the paper would be significantly strengthened. The missing SM and placeholder references are also a completeness issue that must be fixed. The comparison with the authors' prior work [23] is appropriate, and the novelty of the Langmuir-kinetics effects is clear. I recommend major revision rather than rejection because the central derivation is sound within the model and the load-bearing points are addressable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe paper is a real extension of Singh & Proesmans's boundary-driven positional information framework, adding site-dependent Langmuir kinetics and working out Pareto fronts between information I, entropy production Σ, and current J. The exact results are the core value: for the clustered profile, the density solution is simple and the bound Im = log2((1+e)/e) ≈ 0.452 (versus 0.278 with no reactions) is a clean analytic result. The Σ–J fronts, eqs (33) and (41), are derived exactly and they look right. The WKB section for slowly varying enzyme profiles is a sensible addition and the comparison to the Airy-function exact solution is honest.\n\nThe soft spots are concentrated in Section IV B, the uniform-profile dark region. The claimed first-order transition and 'thermodynamically suboptimal zone' depend on a concave segment of the I–Σ Pareto front. The authors obtain it by sweeping λ in Ω = −λI + (1−λ)Σ, eq (34). As you know, linear scalarization only traces the convex hull of a non-convex front; it cannot produce a concave segment. The text infers that segment from optimizer jumps and 'metastable' profiles, and points to eq (??) in a Supplemental Material that is not in the submitted source. That means the central novelty in the uniform case is not independently checkable from what is on the page. This is a genuine weakness, but it is addressable: the cusp value and the coexisting profiles need to be derived or computed directly, and the SM must be supplied. I would not call the whole phase-transition picture invalid; the exact results around it stand, and the hybrid transition language is more speculative than the convex-front claims.\n\nThe reference list is fine; the comparison to [23] is a consistency check, not circularity. The paper does not fit anything to data, so the usual circularity worry does not apply.\n\nFor a reader working on information-thermodynamics trade-offs or on Pareto optimization in stochastic thermodynamics, this is worth engaging with. It deserves a serious referee, but the referee should be asked to verify the missing SM and to re-examine whether the concave segment is real or an artifact of the scalarization. I would not desk-reject it.\n\nRecommendation: send to peer review, with a strong request for the SM and for a direct computation of the cusp/concave branch.","headline":"A mostly clean analytic Pareto-front paper whose strongest new claim—first-order switching in the uniform-profile optimum—rests on a concave segment that the submitted text and missing SM do not yet make checkable.","tokens_in":18559,"tokens_out":1939,"would_cite":true,"duration_ms":19918,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["82C22","82C31","94A17","92C15"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper claims that in boundary-driven exclusion processes with position-dependent Langmuir kinetics, the Pareto-optimal trade-offs among positional information, dissipation, and reaction current are non-trivial, with a clustered…","keywords":["positional information","Pareto-optimal trade-offs","simple exclusion process","Langmuir kinetics","entropy production","reaction-diffusion system","mutual information","phase transitions"],"falsifier":"Enumerate all nondominated $(I,\\Sigma)$ pairs for the uniform profile in the dark phase (for example, $\\gamma=0.1$, $\\alpha=9$) by dense grid search over $(\\rho_L,\\rho_R)$ in $[0,1]^2$. If the concave section near $\\lambda_{c,2}$ contains no point that is truly Pareto-optimal, meaning every candidate there is dominated by another profile, then the first-order transition and the thermodynamically suboptimal zone are not real features of the model.","tokens_in":17581,"feed_emoji":"🧬","tokens_out":12168,"duration_ms":109687,"temperature":0.7,"pith_summary":"The paper tries to establish that in a one-dimensional lattice gas with hard-core particles, connected to particle reservoirs at its two ends and with particle attachment and detachment that varies from site to site, there are sharp trade-offs between how much positional information a steady density profile carries and how much entropy production and reaction current it costs. Its headline quantitative result is that a single sharply clustered reaction site can raise the maximum positional information to $\\log_2((1+e)/e)\\approx 0.452$ bits, reached with maximally asymmetric reservoirs, the reaction driven far forward, and infinitely fast reactions, exceeding the $0.278$-bit ceiling of the same channel without site-dependent reactions. For a uniform reaction-site profile, the optimal boundary protocol undergoes phase-transition-like changes as the optimization weight shifts from minimizing dissipation to maximizing information: a continuous transition in one parameter regime, and in another a coexistence segment followed by a first-order-like jump from monotonic to non-monotonic density profiles. If correct, this provides a concrete thermodynamic picture of developmental positional information: gradients carry only limited bits, and the last bits are disproportionately expensive.","feed_headline":"Reaction sites can push positional information to 0.452 bits","feed_subtitle":"A clustered enzyme beats the purely diffusive limit of 0.278 bits, at the cost of extra dissipation.","key_machinery":"The load-bearing object is the steady-state density equation $\\rho''(\\kappa)=\\alpha^2 e(\\kappa)(\\rho(\\kappa)-\\gamma)$, where $\\alpha^2=(\\omega_{\\rm on}+\\omega_{\\rm off})/p$ is the second Damkohler number measuring reaction relative to diffusion, $\\gamma=\\omega_{\\rm on}/(\\omega_{\\rm on}+\\omega_{\\rm off})$ is the Langmuir isotherm, and $e(\\kappa)$ is the spatial profile of reaction sites. From the solution of this equation one computes the positional information $I$ (mutual information between occupancy and position), the rescaled entropy production $\\Sigma$, and the global reaction current $J$. The search over boundary densities is converted into a one-parameter problem through the scalarized objective $\\Omega=-\\lambda I+(1-\\lambda)\\Sigma$; as $\\lambda$ increases from 0 to 1, the optimizer moves along the Pareto front, and kinks, linear segments, and concave parts of that front are interpreted as phase transitions in the optimal protocol.","core_discovery":"The central claim is that the steady-state density profile of a boundary-driven simple symmetric exclusion process with Langmuir kinetics determines a Pareto front among positional information $I$, rescaled entropy production $\\Sigma$, and global reaction current $J$ that is generically not a simple curve: it has kinks, cusps, convex and concave pieces, and the optimizing boundary densities $(\\rho_L,\\rho_R)$ switch between distinct branches. For the clustered profile $e(\\kappa)=E_T\\delta(\\kappa-\\kappa_0)$ the paper finds the exact upper bound $I_m=\\log_2((1+e)/e)\\approx 0.452$ bits, attained at $\\rho_L=1$, $\\rho_R=0$, $\\gamma\\to 0$, $\\alpha\\to\\infty$, $\\kappa_0=1-\\tanh(1/2)$, exceeding the no-Langmuir value $\\log_2(2/\\sqrt{e})\\approx 0.278$ bits. For the uniform profile, the $I$-$\\Sigma$ front is convex in one region of the $(\\gamma,\\alpha)$ phase diagram, giving a second-order-like transition in the optimal protocol as the weight $\\lambda$ increases, while in the other region the front contains a linear segment where multiple profiles coexist at $\\lambda_{c,1}\\approx 0.728$ and a concave segment near $\\lambda_{c,2}\\approx 0.967$ where the global optimum jumps from monotonic to non-monotonic profiles, with metastable hysteretic profiles and a thermodynamically suboptimal zone.","pith_inferences":["Beyond the paper, the 0.452-bit ceiling implies that a single binary occupancy readout along one axis cannot provide multi-bit positional information; multi-bit readouts would require multiple molecular species, non-binary variables, or multidimensional decoding.","This suggests a direct experimental test: in a microfluidic channel with immobilized enzymes and controllable reservoir densities, cyclically sweeping the boundary densities near the predicted first-order transition should produce hysteresis in the measured density profile.","The exact clustered-profile bound also offers a design benchmark: any proposed enzyme arrangement or active-transport scheme can be compared against $\\approx 0.452$ bits for the same one-bit readout, and operating near the continuous transition $\\lambda_c$ maximizes bits per unit dissipation.","If the concave front is taken at face value, it predicts bistability: for a range of dissipation budgets, monotonic and non-monotonic density profiles are both locally optimal, so small perturbations could flip a system between qualitatively different morphogen profiles."],"forward_implications":["A single localized reaction site can raise the maximum positional information of a one-bit occupancy readout by about 60% relative to a purely diffusive boundary-driven channel (0.452 vs 0.278 bits), so position-dependent reaction kinetics can act as an information resource, not only as a dissipative cost.","On convex parts of the Pareto front, diminishing returns are exact in the sense that moving near the information maximum requires diverging entropy production; the last fractions of a bit are thermodynamically expensive.","In the uniform-profile dark phase, the optimal boundary protocol exhibits phase-coexistence-like behavior: at $\\lambda_{c,1}$ several monotonic profiles yield the same scalarized cost, and at $\\lambda_{c,2}$ the global optimum jumps discontinuously to non-monotonic profiles.","The concave (non-convex) section of the front marks a thermodynamically suboptimal zone: intermediate profiles there are never globally optimal, so a system in that zone can improve both information and dissipation by moving to either bounding branch.","The WKB approximation reproduces the exact trade-offs qualitatively for slowly varying enzyme profiles, so the same kinks and non-convexities should appear for general position-dependent Langmuir kinetics, not only the two exactly solvable profiles."],"supporting_citations":[{"why":"supplies the boundary-driven simple exclusion process with Langmuir kinetics and the 1/N^2 rate scaling that makes the steady-state density equation exact.","marker":"[32]"},{"why":"provides the no-Langmuir baseline maximum $\\log_2(2/\\sqrt{e})\\approx 0.278$ bits and the boundary-driven positional-information setup this paper extends.","marker":"[23]"},{"why":"defines the information-theoretic positional information measure used in equation (10).","marker":"[9]"},{"why":"motivates clustered enzyme profiles and their role in shaping morphogen gradients and reaction flux.","marker":"[37]"},{"why":"supplies the enzyme-arrangement performance context that justifies the clustered profile as the central example.","marker":"[38]"},{"why":"supports the existence of locally non-convex mutual-information versus entropy-production Pareto fronts.","marker":"[52]"},{"why":"supplies the exact solution method for the boundary-driven exclusion process on which the model is built.","marker":"[29]"}],"fun_headline_variants":["Clustered sites hit 0.452-bit positional info ceiling","Boundary tuning beats diffusive limit: 0.452 bits of position","Reaction boundaries tune positional info to 0.452 bits","Clustered enzymes beat diffusive positional limit at 0.452 bits"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that changing the single weight $\\lambda$ in the combined objective $\\Omega=-\\lambda I+(1-\\lambda)\\Sigma$ visits every genuinely optimal trade-off between information and dissipation; if the curved-inward segment of the trade-off curve is missed or is an artifact of the search, the claimed abrupt switch in optimal protocol and the 'suboptimal zone' would not be established.","fun_headline_variants_meta":{"raw":{"variants":["Clustered sites hit 0.452-bit positional info ceiling","Boundary tuning beats diffusive limit: 0.452 bits of position","Reaction boundaries tune positional info to 0.452 bits","Clustered enzymes beat diffusive positional limit at 0.452 bits"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000806,"raw_usage":{"total_tokens":3580,"prompt_tokens":1027,"completion_tokens":2553,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":643,"completion_tokens_details":{"reasoning_tokens":2475}},"tokens_in":643,"tokens_out":2553,"duration_ms":19946,"temperature":1.0,"reasoning_tokens":2475,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T23:03:09.651188+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Enumerate all nondominated $(I,\\Sigma)$ pairs for the uniform profile in the dark phase (for example, $\\gamma=0.1$, $\\alpha=9$) by dense grid search over $(\\rho_L,\\rho_R)$ in $[0,1]^2$. If the concave section near $\\lambda_{c,2}$ contains no point that is truly Pareto-optimal, meaning every candidate there is dominated by another profile, then the first-order transition and the thermodynamically suboptimal zone are not real features of the model.","supporting_citations":[{"cited_title":"Derrida, J","cited_arxiv_id":null,"evidence_quote":"supplies the boundary-driven simple exclusion process with Langmuir kinetics and the 1/N^2 rate scaling that makes the steady-state density equation exact."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"provides the no-Langmuir baseline maximum $\\log_2(2/\\sqrt{e})\\approx 0.278$ bits and the boundary-driven positional-information setup this paper extends."},{"cited_title":"Dupin, L","cited_arxiv_id":null,"evidence_quote":"defines the information-theoretic positional information measure used in equation (10)."},{"cited_title":"Grieneisen, B","cited_arxiv_id":null,"evidence_quote":"motivates clustered enzyme profiles and their role in shaping morphogen gradients and reaction flux."},{"cited_title":"Buchner, F","cited_arxiv_id":null,"evidence_quote":"supplies the enzyme-arrangement performance context that justifies the clustered profile as the central example."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"supports the existence of locally non-convex mutual-information versus entropy-production Pareto fronts."}],"review_version":1}