{"id":"c35ede41-f010-4682-bf3d-1b6379d29618","arxiv_id":"2504.16539","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Motility-dependent chemical deposition with saturating chemotactic sensitivity produces a re-entrant phase diagram including fractal-like networks, and a continuum linear stability analysis is used to explain the boundaries.","lead":"A computer model shows that particles that leave a chemical trail only while moving, and are attracted to that chemical, spontaneously form branching, fractal-like networks, and even return to a spread-out state when the chemical builds up too strongly. The result offers a simple physical explanation for how networks in living tissues, such as early blood vessels, might form without a central blueprint.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The continuum LSA is not an independent confirmation: the key parameter ρ_c is fitted to the same simulations and takes inconsistent values, so Eq. 3 may match the phase diagram by construction.","rationale":"The agent-based simulation results are the strongest part of the paper: the bimodal cluster-size distributions, giant number fluctuations, quench growth laws, and the re-entrant morphologies are self-consistent phenomenological evidence. The load-bearing weak point is the claimed theoretical confirmation. For Eq. 3 to confirm the phase diagram, the deposition term αρ(ρ_c − ρ) must be a faithful and preferably parameter-independent coarse-grained representation of the microscopic rule that only moving agents deposit. The inconsistent fitted ρ_c values across SI Fig. S2 and Fig. 3 show that this is not the case, and the paper's own admission that the simulated phase boundary is α-independent while Eq. 3 is α-dependent reinforces the concern. This does not refute the simulations, but it removes the independent support for the statement that the continuum model confirms the transitions. The proposed test settles the issue directly: an independently measured ρ_c that is constant across parameters and reproduces the simulated boundaries would validate the LSA; otherwise the theoretical confirmation should be weakened. Therefore the appropriate verdict remains conditional, and no change from the reader's verdict is needed.","tokens_in":16621,"tokens_out":5109,"duration_ms":50012,"concrete_test":"Measure P_acc(ρ), the probability that a trial move is accepted, as a function of coarse-grained local density ρ in the agent-based model with chemotactic coupling turned off (α = 0) but WCA interactions on. Since deposition occurs only on accepted moves, the microscopic deposition rate is αρP_acc(ρ); fit it to αρ(1 − ρ/ρ_c) and extract ρ_c from this independent measurement. Then insert the measured ρ_c into Eq. 3 and recompute the instability window for α = 5 and α = 100 at β = 0.008. If the predicted boundaries do not match the simulated HP/SN and SC/RHP transitions, or if the independently measured ρ_c differs substantially from the SI Fig. S2 and Fig. 3 fitted values, the LSA confirmation is not genuine. Also check whether ρ_c from this bare measurement is constant across α and β.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Eq. 2 replaces the microscopic motility-dependent deposition rule with αρ(ρ_c − ρ). The functional form is not derived; it is inferred from steady-state c-versus-ρ data in SI Fig. S2, and ρ_c = a/b is a fit parameter taken from the same simulations that Eq. 3 is supposed to explain. Because ρ_c enters χ′ and hence the predicted instability window, the agreement between Eq. 3 and the phase diagram is partly circular. The fit is also not stable: SI Fig. S2 reports ρ_c = 1.47 for the DN phase at α = 5, β = 0.008 and ρ_c = 1.07 for the SC phase at α = 100, β = 0.008, whereas the quench-profile fits in Fig. 3 give ρ_c = 1.002 for DN at α = 5, β = 0.008 and ρ_c = 2.712 for SC at α = 100, β = 0.008. If ρ_c were a motility or packing threshold of the WCA system, it should be approximately independent of α, β, and of whether one fits a cluster edge or a global steady state. The paper itself notes that the simulated HP/SN and SC/RHP boundaries are independent of α, unlike Eq. 3. A fitted, phase-dependent ρ_c is exactly the kind of compensation that could hide this discrepancy. Thus the paper's central theoretical claim—that linear stability criteria determine the observed phases—is not independently established.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies a minimal agent-based model of self-organized pattern formation in which motile agents deposit a chemical only while moving and bias their motion along local chemical gradients with a saturating response. The authors map out a phase diagram in the deposition-decay parameter space (alpha, beta), identifying homogeneous (HP), sparse network (SN), dense network (DN), system-spanning cluster (SC), and a re-entrant homogeneous phase (RHP) at very low decay rates. A continuum model for the density and chemical fields is introduced, with a motility-dependent deposition term alpha*rho*(rho_c - rho), and a linear stability analysis yields an instability window in alpha/beta (Eq. 3). The paper further reports quench dynamics, fractal dimension measurements, cluster-size distributions, mean-squared displacement, and a perturbation study to support the phase classification. The biological discussion relates the networks to vasculogenesis and epithelial dispersal.","tokens_in":16954,"tokens_out":3152,"duration_ms":33754,"significance":"If the theoretical confirmation held up, this would be a valuable contribution: the model is simple, the phase diagram is rich, and the re-entrant homogeneous phase caused by sensitivity saturation is a genuinely interesting prediction that could be tested in artificial active-matter or cell-culture systems. The simulation evidence for the phase diagram, including fractal dimension, system-spanning statistics, and the separate quench and perturbation studies, is substantial and mostly self-consistent. The main weakness is that the continuum linear stability analysis, which is advertised as confirming the transitions, depends on a parameter rho_c fitted to the same simulation data and also fails to reproduce the alpha-independence of the simulated phase boundaries. The central theoretical claim is therefore not independently established in the current version.","major_comments":[{"comment":"The linear stability condition Eq. 3 is not an independent confirmation of the phase diagram because rho_c is obtained by fitting the steady-state relation c(rho) = a*rho - b*rho^2 to the very same agent-based simulations that Eq. 3 is supposed to explain. The fit is not stable across contexts: SI Fig. S2 reports rho_c = 1.47 for the DN phase at alpha = 5, beta = 0.008 and rho_c = 1.07 for the SC phase at alpha = 100, beta = 0.008, whereas the quench-profile fits in Fig. 3 give rho_c = 1.002 for DN at alpha = 5, beta = 0.008 and rho_c = 2.712 for SC at alpha = 100, beta = 0.008. If rho_c represented a motility or packing threshold of the underlying WCA system, it should be approximately independent of alpha, beta, and of whether one fits a global steady state or a cluster interface. The authors should either derive rho_c from the microscopic deposition rule (e.g., from the probability that a trial move is accepted as a function of local density) or demonstrate that the predicted instability boundary is insensitive to the fitted value over the observed spread; otherwise Eq. 3 should be presented as a consistency check rather than as confirmation.","section":"Section II D, Eq. 3 vs. Fig. 2(a)"},{"comment":"The paper itself acknowledges that the simulated HP/SN and SC/RHP boundaries are independent of alpha, whereas Eq. 3 depends on alpha/beta through c0. Since Fig. 2(a) draws the solid line from Eq. 3, the figure claims a quantitative match that the text immediately qualifies. This mismatch is load-bearing for the statement that linear stability criteria determine the observed phases. A quantitative comparison is needed: for example, plot the simulated boundaries in the (alpha/beta)-rho0 plane and compare with the theoretical C_- and C_+ contours for fixed rho_c, or show that the alpha-independence emerges in a suitable limit (e.g., when the instability window is wide in alpha/beta). As written, the theoretical boundary does not predict the alpha-dependence of the most prominent transitions.","section":"Section II D and SI Eq. S1"},{"comment":"The continuum model in Eq. 2 is posited rather than derived from the microscopic rules. In particular, the deposition term alpha*rho*(rho_c - rho) is justified only by the steady-state fit in SI Fig. S2, and the effective parameters D, chi, and rho_c are not measured independently from the agent dynamics. The linear stability analysis then becomes a self-consistency argument: the assumed deposition nonlinearity is tuned to reproduce the simulation data, and the instability boundary inherits that tuning. The authors should either (i) coarse-grain the microscopic update rules to derive the deposition term and the effective diffusivity and chemotactic mobility, or (ii) explicitly state that Eq. 2 is a phenomenological model and validate it by testing its predictions (not just its stability boundary) against independent simulation observables, such as front propagation speeds or density profiles.","section":"Section II E and SI Fig. S1"},{"comment":"The quench-dynamics interpretation in Section II E relies on comparing c(x) = rho(x)[rho_c - rho(x)] with rho_c > 2*rho_i versus rho_c < 2*rho_i to distinguish the SC and DN regimes. However, the rho_c values used for these two cases come from the same fitted functional form, and the figures show only single representative fits (Fig. 3(e) and 3(j)). To support the claim that this criterion predicts whether the network instability is present, the authors should show that the inequality rho_c > 2*rho_i robustly separates the SC and DN regions across a range of alpha and beta, and report the error bars on rho_c in these fits.","section":"Section II D, Eq. 1"},{"comment":"Equation (1) introduces the deposition rule through the function h({r, r_dot}), but the connection between this microscopic rule and the continuum closure alpha*rho*(rho_c - rho) is never made explicit. The sentence 'The functional dependence rho(rho - rho_c) captures the decrease in the rate of chemical deposition with reduced motility at high density controlled by rho_c' is not a derivation. Please clarify whether h is meant to be the occupancy-dependent acceptance probability and how the linear-in-density form emerges from it.","section":"Figure 2(a) caption"},{"comment":"The caption of Fig. 2(a) refers to the solid line as 'a linear stability boundary obtained using Eq. 3' but does not list the parameter values (rho_c, chi, D, rho_0) used to generate it. Since those parameters are not all fixed by the simulation protocol, the reader cannot check whether the drawn boundary is representative or tuned. Please provide the parameter values and, ideally, a shaded region showing the full unstable band C_- < c0 < C_+ rather than a single contour.","section":null}],"minor_comments":[{"comment":"The terminology alternates between 'concentration-sensitive feedback', 'saturating sensitivity', and 'concentration-limited responsiveness'; please use one consistent term throughout.","section":"Abstract and Introduction"},{"comment":"The caption reports (a,b) and rho_c = a/b, but it does not mention that these values differ substantially from the rho_c values used in the main-text Fig. 3 fits for the same nominal phases and parameters; explicitly flagging this would improve transparency.","section":"SI Fig. S2 caption"},{"comment":"The description of the re-entrant phase says particles become 'effectively insensitive to the uniform chemical field'; since the field is uniform in RHP, it would be clearer to state that the gradient is too weak to bias motion even though the concentration is high.","section":"Section II A"},{"comment":"The time step is described with r(t+1) but the model is said to be continuous in time; please clarify the discrete-time update and the relation between v0*Delta_t = 1 and the WCA diameter sigma.","section":"Methods"},{"comment":"The growth-law fits (z = 3 for SC and exponential for DN) are presented without confidence intervals or goodness-of-fit measures; adding these would strengthen the claim of two distinct coarsening regimes.","section":"SI Section VI"},{"comment":"Reference [1] currently appears to have an incomplete author list ('J.-L. Deneubourg, N. R. Franks, J. Sneyd, G. Theraulaz, E. S. C. ... Camazine'); please correct all bibliographic entries for consistency.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The simulation phase diagram and the fractal characterization are the strongest parts of the paper. The central problem is the status of Eq. 3: as written, it is a consistency check with a fitted parameter, not an independent prediction, and the paper admits a qualitative mismatch with the alpha-independence of the simulated boundaries. I would recommend major revision rather than rejection because the authors can plausibly fix this by deriving rho_c from the microscopic update rules, by performing an independent parameter-free comparison, or by explicitly reframing the LSA as a posterior rationalization. I would also encourage the editor to ask the authors to make the parameter values used to draw the theoretical boundary in Fig. 2(a) available, as this is currently a reproducibility gap."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The thing you should know: the simulations are probably fine, but the continuum theory is not an independent confirmation. The central result—a re-entrant phase diagram where lower decay first drives network formation, then full clustering, then back to homogeneity—is a nice minimal finding. Deposition only while moving plus saturating chemotactic response is a plausible pair of ingredients, and the authors do a good job characterizing phases with cluster size distributions, fractal dimension, and quench dynamics.\n\nThe soft spot is Section II.D and Eq. 3. The continuum model replaces the microscopic deposition rule with αρ(ρ_c − ρ), and ρ_c is not derived. It is fit to steady-state c-versus-ρ curves from the same simulations (SI Fig. S2), so the linear stability analysis agreeing with the phase diagram is partly circular. Worse, the fitted ρ_c is not stable: SI gives ρ_c = 1.47 for a DN state and 1.07 for an SC state, while the quench-profile fits in Fig. 3 give 1.002 and 2.712 for the same parameter points. If ρ_c were a motility or packing threshold, it should not vary that much. The paper also admits the simulated phase boundary is independent of α, unlike Eq. 3. So the theory is doing less work than it claims.\n\nThis does not sink the simulation work. The mechanism is clear, the biological analogies are appropriately speculative, and the fractal-network observation is worth having. But the authors should either derive ρ_c from the microscopic rules, measure it independently, or reframe the LSA as a qualitative rationalization rather than a confirmation. Minor issues: phase boundaries are visual guides with no error bars, and no code or data are deposited beyond the SI.\n\nI would send this to peer review. A serious referee should push on the circularity and the α-dependence mismatch. If those are fixed, or the theory claims are dialed back, the paper is publishable. It deserves referee time, not a desk reject.","headline":"The simulation phase diagram is likely the contribution; the continuum theory does not independently confirm it because the key threshold is fit to the same simulations.","tokens_in":17462,"tokens_out":3364,"would_cite":true,"duration_ms":34710,"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 minimal two-rule model of self-chemotaxis produces fractal-like networks and then re-enters a uniform state at very low chemical decay.","keywords":["self-organization","chemotaxis","fractal networks","agent-based model","re-entrant phase transition","motility-dependent deposition","linear stability analysis","vasculogenesis"],"falsifier":"Run the agent-based model at fixed $\\beta$ while scanning $\\alpha$ over at least two decades and locate the homogeneous-to-network and network-to-homogeneous boundaries from the cluster-size distribution; Eq. (3) says those boundaries shift with $\\alpha$, whereas the paper reports them as independent of $\\alpha$, so a boundary that moves with $\\alpha$ would falsify the stated linear-stability confirmation.","tokens_in":16430,"feed_emoji":"🕸️","tokens_out":12756,"duration_ms":113575,"temperature":0.7,"pith_summary":"This paper argues that two local rules—agents deposit a chemical only while moving, and their motion is biased up the chemical gradient with a sensitivity that saturates at high concentration—are enough to generate the branching, fractal-like networks seen in early blood-vessel formation and in dispersing epithelial cells. As the chemical decay rate is lowered, the model produces a re-entrant sequence: uniform, sparse network, dense fractal network, compact cluster, and then uniform again, because at extremely low decay the chemical field becomes so strong that the agents' response saturates and they stop following it. The authors back this phase diagram with a continuum model whose linear stability analysis gives a finite window of deposition-to-decay ratio for pattern formation, and they show that the cluster and network phases grow by different mechanisms. If correct, the paper supplies a physically grounded mechanism that could be tested in cell cultures and implemented in synthetic swarms.","feed_headline":"Two local rules build fractal cell networks","feed_subtitle":"Movers deposit chemicals and follow saturating gradients; very slow decay makes the pattern vanish again.","key_machinery":"The load-bearing object is the feedback loop itself, written in continuum form as coupled equations for particle density $\\rho$ and chemical concentration $c$: $\\dot{\\rho}=-\\nabla\\cdot[\\chi\\rho f'(c)\\nabla c-D\\nabla\\rho]$ and $\\dot{c}=\\alpha\\rho(\\rho_c-\\rho)-\\beta c$, with $f(c)=c/(1+c)$. The deposition term $\\alpha\\rho(\\rho_c-\\rho)$ encodes the microscopic rule that only moving agents deposit: the source is quadratic in density, vanishes at density $\\rho_c$, and therefore peaks at intermediate density, so the interior of a cluster stops producing chemical while its edges continue. Linearizing around the homogeneous state gives the instability condition $C_-