{"id":"96948c37-61e5-49b8-95ff-58a66d0f738d","arxiv_id":"2411.11696","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"In coarse-grained simulations, the number and steady-state size of protein droplets are selected by the concentration and diffusivity of two antagonistic enzyme populations, because growth is interrupted by encounters with the dispersing enzyme.","lead":"This computational study models droplets that form from a two-state protein, where two types of moving enzymes convert the protein between a droplet-forming state and a dispersed state. It finds that enzyme concentration and how fast enzymes move control how many droplets form and what size they reach, which is relevant to understanding membraneless compartments in cells.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Enzyme self-propulsion at droplet interfaces, acknowledged in the paper, may make the imposed diffusion coefficient D an unreliable control variable; the quantitative claim that D alone sets droplet size is not yet established.","rationale":"The reader's weakest assumption correctly identifies enzyme self-propulsion as the most serious threat to the quantitative claim. The paper's own text admits that in the hybrid method self-propulsion may dominate normal diffusion, and even in full BD the D = 0.2D0 fit improves if a slightly larger diffusion coefficient is used. That admission means the model's control parameter is not cleanly the imposed D. The two simulation methods agreeing qualitatively is real support for a concentration-dependent size selection, but the specifically quantitative statement about enzyme diffusivity governing size depends on the encounter-time expression Eq. 5, whose only free parameters are fitted and whose transport coefficient is not independently verified. The proposed MSD test is a direct, low-cost check: if Deff ≈ D, the concern does not land and the diffusivity claim is supported; if Deff deviates, the paper must either use Deff as the control variable or reframe the claim as qualitatively valid only. I do not see a reason to reject the paper; the existing CONDITIONAL verdict remains appropriate, and this check would clarify whether the condition is actually necessary.","tokens_in":14830,"tokens_out":5446,"duration_ms":56739,"concrete_test":"In the full BD simulations, measure the mean-squared displacement of EB→A enzymes separately for enzymes located within rcut of a droplet interface and those in the bulk A phase, for each imposed enzyme D (especially D = 0.2D0 and 5D0). Extract the long-time effective diffusion coefficient Deff from the slope of MSD/(4t). If Deff differs from the imposed D by more than ~20% for the interface-associated population, or if Deff varies with droplet size or ρB→A, then the control variable is an emergent effective motility, not the bare D; Eq. 5 and the associated fits in Fig. 6 would need to be re-evaluated using Deff before the claim that diffusivity selects droplet size can be sustained.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Central claim: enzyme concentration and diffusion coefficient govern droplet size (Eq. 4). The D-dependence rests on Eq. 5, which assumes te ∝ 1/(DEB→A + Ddp), i.e., enzyme transport is ideal Brownian with bare D. The paper itself reports in 'Enzyme diffusivity affects the size of condensates' that in full BD at D = 0.2D0 the fit improves with a slightly higher diffusion coefficient, and that MSD 'show a slight increase of the diffusion coefficient'; for the HM, 'a first investigation of the mean squared displacements of the enzymes suggests a strong self-propulsion in this range of parameters, which may dominate normal diffusion.' Thus the actual transport coefficient of enzymes is not the imposed D but an emergent, interface-driven motility that is neither predicted nor controlled. If this effective motility depends on droplet size, enzyme density, or local composition, then the relation between the simulation parameter D and droplet size is not a clean cause-and-effect; the analytic model absorbs this through fitted α and β and the self-consistent Ddp = D0/⟨N⟩∞ (Eq. 7), so Eq. 4 does not independently validate the encounter-time mechanism. The HM regime, where self-propulsion dominates, is explicitly not captured by Eq. 4. This is the load-bearing weakness: the paper's headline quantitative claim about diffusivity control is not backed by a measured effective diffusivity.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"Fries et al. propose a two-dimensional model of biocondensate formation in which two enzyme species catalyze opposing interconversions of a two-state protein (A and B), with B proteins attracting each other and forming droplets. Enzymes are explicit Brownian particles, and reactions occur only within a cutoff radius of an enzyme. The authors use two complementary simulation methods: all-particle Brownian dynamics (full BD) and a hybrid method that couples Cahn-Hilliard-Cook dynamics for the protein field with Brownian dynamics for the enzymes (HM). They report that adding EB→A enzymes arrests Ostwald ripening, producing finite steady-state droplet sizes that decrease with increasing EB→A concentration and with increasing enzyme diffusivity, and a non-monotonic number of droplets per EA→B enzyme as a function of EB→A concentration. An analytic model (Eq. 4) with fitted constants reproduces the full BD data for the mean droplet size evolution and steady-state size, and the authors interpret this as evidence that the encounter time between droplets and EB→A enzymes controls droplet size.","tokens_in":15175,"tokens_out":6849,"duration_ms":64829,"significance":"If the conclusions hold, the paper offers a minimal mechanistic picture of how enzyme concentration and transport can select biocondensate size, with explicit enzyme trajectories being a methodological advantage over density-dependent reaction-rate models. The internal agreement between two independent simulation methods for the main qualitative trends (arrested growth, finite steady-state sizes, smaller droplets with faster enzymes, non-monotonic droplet number) is a genuine strength. However, the quantitative status of the central claim is currently limited by the fitted analytic model and by the paper's own observation of interface-induced enzyme self-propulsion, which may contradict the assumption that the imposed diffusion coefficient D is the relevant transport control parameter.","major_comments":[{"comment":"The claim that the bare diffusion coefficient D of enzymes controls droplet size is not fully established because the paper reports that enzyme transport near droplets is not ideal Brownian. In the section 'Enzyme diffusivity affects the size of condensates', the authors state that for D = 0.2D0 the fit improves with a slightly higher diffusion coefficient, that MSD 'show a slight increase of the diffusion coefficient' in full BD, and that in the HM 'a first investigation of the mean squared displacements of the enzymes suggests a strong self-propulsion in this range of parameters, which may dominate normal diffusion.' Equations 4-5, however, assume te ∝ 1/(D_EB→A + Ddp), i.e., transport governed by the imposed D. If the effective enzyme mobility is an emergent, interface-driven quantity that is not equal to D, then Fig. 6 does not demonstrate that D alone selects droplet size; the control variable would be an unexplained effective motility. The authors should measure and report effective diffusion coefficients from MSD in both simulation methods and show how they scale with D, or qualify the claim to refer to effective motility rather than the bare diffusion coefficient.","section":"Enzyme diffusivity affects the size of condensates; Eq. 5"},{"comment":"The analytic model is fitted, not independently predictive. The constants α = 2.7 and β = 900 are fit parameters, and Eq. 7 sets Ddp = D0/⟨N⟩∞, where ⟨N⟩∞ is the measured steady-state droplet size. Because the predicted steady-state size appears in the definition of the droplet diffusivity used in the prediction, the agreement of Eq. 4 with simulation is a consistency check of a two-parameter functional form rather than a validation of the encounter-time mechanism. The paper should state the predictive content explicitly: with α and β fixed from one dataset, does the implicit equation for ⟨N⟩∞ (via Eqs. 4-7) predict the full dependence on ρB→A and D? If so, show that test.","section":"Appendix A3, Eqs. 4-7"},{"comment":"The manuscript does not provide convergence tests for the claimed stationary states or error bars for the plotted means, despite stating that results are averaged over 50 independent realizations. For example, the caption of Fig. 3 says 'the stationary state is assumed to be reached in the interval [1800, 2400] Dt/σ²' without comparing adjacent time windows or showing the time series of the mean and variance. Since the paper makes quantitative claims, including the non-monotonic dependence of droplet number on ρB→A and the dependence on D, the absence of uncertainty quantification makes it difficult to assess the significance of the reported differences and the quality of the fits.","section":"Figs. 3, 4, 6; stationary-state definitions"}],"minor_comments":[{"comment":"Typographical errors: 'explicitely' should be 'explicitly', and 'Theses reactions' should be 'These reactions'.","section":"Models"},{"comment":"In Eq. 24, the term c dt²/τ is second order in dt and is neglected without comment; please explicitly state this approximation.","section":"Appendix A3, Eq. 24"},{"comment":"The text contains 'simularions' for 'simulations'.","section":"Appendix A1e"},{"comment":"Full BD and HM are simulated in different parameter regimes (e.g., ρA→B = 1.6×10⁻³ in full BD versus 1.2×10⁻⁴ in HM). The paper should state explicitly that the comparison between the two methods is only qualitative and that quantitative agreement is not expected.","section":"Figs. 3-6"},{"comment":"The droplet detection criterion is described as 'The square root of this threshold, 2σ, is taken as the distance criteria'; this sentence is unclear and should be rewritten.","section":"Appendix A1a"}],"recommendation":"major_revision","confidential_remarks":"The qualitative conclusions are plausible and the two-method cross-check is a strength. The main risk is that the quantitative claim about diffusivity control rests on an assumption (ideal Brownian enzyme transport) that the authors themselves show to be violated near droplet interfaces, and on a fitted analytic model that uses the measured steady-state size through Eq. 7. I would ask the editor to require an effective-diffusivity analysis and a clear statement of the analytic model's predictive status before publication. This is not a novelty or scope concern."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is a genuine modeling step, worth a serious look. The authors replace the usual density-dependent reaction rates with explicitly diffusing enzymes that catalyze A↔B conversions locally, and they back it up with two independent simulation schemes. The main message—enzyme concentration and bare diffusivity select droplet size and number by controlling encounter times—is largely supported by the data.\n\nWhat's new: explicit enzyme trajectories in both full Brownian dynamics and a hybrid Cahn-Hilliard-Cook + BD method. That's a real extension of the mean-field active-droplet literature. The two methods agree on the qualitative trends: EB→A enzymes arrest Ostwald ripening, droplet size decreases with EB→A concentration and with enzyme diffusivity, and the number of droplets per EA→B enzyme is non-monotonic. They also run sensible controls (reverse reactions, box size, rcut) and are honest about the hybrid method's quantitative disagreement with the analytic model.\n\nSoft spots: Eq. 4 is not a prediction. The encounter-time model has fitted constants α and β, and the droplet diffusion coefficient Ddp is set to D0/⟨N⟩∞ using the measured steady-state size, so the fit doesn't validate the mechanism independently. There are no error bars on the figures, though results are averaged over 50 realizations. Stationarity is assumed on fixed time windows without convergence checks. Everything is 2D, which matters for droplet coalescence and ripening.\n\nThe more substantive concern is the self-propulsion the authors themselves report. For D=0.2D0, MSDs show enhanced diffusion; in the hybrid method, self-propulsion at interfaces may dominate normal diffusion. That means the imposed D is not necessarily the transport coefficient that sets encounter rates. The qualitative trend with D survives, but the quantitative claim that bare diffusivity alone controls droplet size is less clean than the abstract implies. The paper flags this and doesn't hide it—credit for that—but a referee should ask for effective diffusivity measurements or a discussion of how self-propulsion scales with D.\n\nBottom line: a useful, honest simulation study for the biocondensate/active matter community. It deserves peer review, not a desk rejection. I'd recommend the editor send it out, with the expectation of revisions on error bars, stationarity, and the effective-diffusivity question.","headline":"Genuine modeling step with explicit enzyme trajectories; the central trend holds, but the analytic model is a fit and the D-control claim is softened by enzyme self-propulsion.","tokens_in":15656,"tokens_out":3098,"would_cite":true,"duration_ms":30571,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A two-enzyme system can arrest droplet coarsening, with enzyme concentration and mobility setting the final droplet size.","keywords":["biocondensates","liquid-liquid phase separation","Ostwald ripening","enzymatic reactions","two-state protein","active droplets","explicit enzyme trajectories","encounter time"],"falsifier":"Make an in vitro system with two opposing enzymes acting on a dropletable protein, then measure the steady-state droplet radius as the dispersion-promoting enzyme's diffusion coefficient is varied (for instance by changing solution viscosity). The model predicts that droplet radius decreases monotonically with that diffusion coefficient; a flat or non-monotonic response would contradict the encounter-time mechanism. A second check is to record droplet lifetimes: under the model, they should be exponentially distributed with mean $t_e/(S \\rho_{B\\to A})$.","tokens_in":14649,"feed_emoji":"🧫","tokens_out":8772,"duration_ms":81288,"temperature":0.7,"pith_summary":"The paper claims that a pair of enzymes acting on a protein that can either form droplets or stay dissolved is enough to stop droplet coarsening and select a definite droplet size. It models the protein in a droplet-forming state and a dispersed state, with one enzyme converting the protein into the droplet state and the other converting it back, but only in the enzyme's immediate vicinity; enzyme trajectories are explicit Brownian walks, so reaction rates are not uniform in space. Simulations with two methods—full Brownian dynamics of all particles, and a hybrid that couples a Cahn-Hilliard-Cook diffusion equation to Brownian enzymes—show that a nonzero concentration of the dispersion-promoting enzyme arrests Ostwald ripening, that the steady-state droplet size decreases as this enzyme's concentration or its diffusion coefficient increases, and that the number of droplets rises then falls as that concentration increases. A simple analytic expression, $\\langle N \\rangle = (c t_e / S \\rho_{B\\to A})(1 - \\exp(-t S \\rho_{B\\to A}/t_e))$, captures the droplet growth data, tying the steady-state size to the encounter time between a droplet and the dispersion enzyme. The reason to care: this gives cells a minimal physical route—enzyme abundance and mobility, not just reaction chemistry—for controlling the size and number of membraneless compartments.","feed_headline":"Opposing enzymes can freeze droplet ripening at a finite size","feed_subtitle":"Enzyme concentration and mobility set the size droplets reach before being dissolved.","key_machinery":"The load-bearing object is the encounter-time model of droplet destruction. The paper assumes each growing droplet receives a constant protein influx $c$, that droplets are destroyed instantly upon meeting an $E_{B\\to A}$ enzyme, and that encounter times are exponentially distributed with mean $t_e$. This reduces the mean droplet size to an exponential-saturation formula, Eq. 4, whose steady-state value is $c t_e/(S \\rho_{B\\to A})$. The second piece of machinery is the explicit Brownian representation of enzymes, which replaces the artificial density-dependent reaction rates used in previous mean-field droplet models with reaction rates that emerge from enzyme trajectories, and in the hybrid method couples those trajectories to a Cahn-Hilliard-Cook field.","core_discovery":"In a model where two enzyme species catalyze opposite conversions between a condensate-prone protein state B and a dispersed state A, the paper's central discovery is that the stochastic, spatially localized action of the enzymes arrests the otherwise uninterrupted growth of droplets. With only the droplet-promoting enzyme present, a single B-rich droplet grows without bound by Ostwald ripening and coalescence. Adding the dispersion-promoting enzyme $E_{B\\to A}$ produces a steady state with coexisting droplets; the mean droplet size falls as $\\rho_{B\\to A}$ increases, and the mean number of droplets per $E_{A\\to B}$ enzyme goes through a maximum. Varying the diffusion coefficient of either enzyme changes droplet size the same way: faster enzymes give smaller droplets. The full-Brownian-dynamics data are described by an encounter-destruction picture in which each droplet grows at constant influx and is abruptly emptied when it meets an $E_{B\\to A}$ enzyme, yielding the analytic form $\\langle N \\rangle = (c t_e/S \\rho_{B\\to A})(1 - \\exp(-t S \\rho_{B\\to A}/t_e))$. The hybrid method, though it shows quantitative differences at large enzyme concentrations, reproduces the same qualitative dependence on enzyme concentration and diffusivity.","pith_inferences":["If the hint of interface-driven enzyme self-propulsion is confirmed, then an enzyme's effective encounter rate with a condensate is set partly by the condensate itself, so enzyme diffusivity should be treated as a coupled variable rather than a fixed input.","The encounter-destruction picture implies droplet lifetimes should be exponentially distributed; recording droplet birth-death histories in experiments or simulations would test this renewal mechanism against models where droplets shrink continuously.","The same machinery could be applied to multiple enzyme species or spatially patterned enzymes, predicting that localizing dispersion enzymes near nucleation sites would create gradients of droplet size across a cell.","Extending the model to three dimensions with hydrodynamic interactions may change the encounter-time scaling, since the diffusion coefficient of a droplet depends on solvent viscosity and droplet size differently beyond two dimensions."],"forward_implications":["In a living cell, raising the expression level of the dispersion-promoting enzyme would shrink condensates continuously, rather than dissolving them abruptly, as long as the droplet-promoting enzyme remains active.","Tuning enzyme mobility—through viscosity, crowding, or active transport—should provide an independent control over condensate size, separate from changing reaction rates.","The non-monotonic dependence of droplet number on enzyme ratio means cells can maximize organelle count by balancing the two enzyme activities at an intermediate value.","The analytic formula predicts that at fixed influx the steady-state droplet size is inversely proportional to the dispersion-enzyme surface concentration, a directly testable scaling law."],"supporting_citations":[{"why":"Review of chemically active droplet models whose density-dependent reaction rates the paper's explicit-enzyme approach avoids.","marker":"[4]"},{"why":"Cites experimental observation that condensate size is arrested in vivo, motivating the search for a non-equilibrium size-selection mechanism.","marker":"[5]"},{"why":"A mean-field model of chemically active droplets that predicts multiple coexisting droplets; this paper extends it by resolving enzyme trajectories.","marker":"[7]"},{"why":"Biological example of two opposing enzymes controlling P-granule dynamics, the motivating case.","marker":"[12]"},{"why":"Evidence and theories that enzyme catalytic activity can change enzyme diffusivity, motivating the study of the diffusion coefficient as a control parameter.","marker":"[27–29]"},{"why":"The authors' earlier finding of a particle propelled by a Lennard-Jones domain, used to interpret enzyme self-propulsion at droplet interfaces.","marker":"[24]"}],"fun_headline_variants":["Stochastic enzyme encounters cap droplet growth","Enzyme collisions freeze droplet ripening at finite size","Enzyme mobility dictates the size of biocondensates","Droplets grow until an enzyme meets and empties them"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument assumes an enzyme's bare diffusion coefficient fully controls how often it meets a droplet, yet the simulations show signs that droplet interfaces may actively propel enzymes, which would change the effective encounter rate.","fun_headline_variants_meta":{"raw":{"variants":["Stochastic enzyme encounters cap droplet growth","Enzyme collisions freeze droplet ripening at finite size","Enzyme mobility dictates the size of biocondensates","Droplets grow until an enzyme meets and empties them"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000357,"raw_usage":{"total_tokens":1957,"prompt_tokens":987,"completion_tokens":970,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":603,"completion_tokens_details":{"reasoning_tokens":907}},"tokens_in":603,"tokens_out":970,"duration_ms":7763,"temperature":1.0,"reasoning_tokens":907,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T18:13:54.896961+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Make an in vitro system with two opposing enzymes acting on a dropletable protein, then measure the steady-state droplet radius as the dispersion-promoting enzyme's diffusion coefficient is varied (for instance by changing solution viscosity). The model predicts that droplet radius decreases monotonically with that diffusion coefficient; a flat or non-monotonic response would contradict the encounter-time mechanism. A second check is to record droplet lifetimes: under the model, they should be exponentially distributed with mean $t_e/(S \\rho_{B\\to A})$.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Cites experimental observation that condensate size is arrested in vivo, motivating the search for a non-equilibrium size-selection mechanism."},{"cited_title":"Zwicker, R","cited_arxiv_id":null,"evidence_quote":"A mean-field model of chemically active droplets that predicts multiple coexisting droplets; this paper extends it by resolving enzyme trajectories."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Biological example of two opposing enzymes controlling P-granule dynamics, the motivating case."},{"cited_title":"Decayeux, V","cited_arxiv_id":null,"evidence_quote":"The authors' earlier finding of a particle propelled by a Lennard-Jones domain, used to interpret enzyme self-propulsion at droplet interfaces."}],"review_version":1}