{"id":"4f535aa1-ff4a-4701-85fc-4715903bbfdc","arxiv_id":"1908.04253","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":15,"one_line_summary":"A reaction-diffusion model with dynamic boundary conditions and wnt-gradient sensing reproduces polarity-preserving planarian regeneration in cutting, grafting, and growth simulations.","lead":"This paper introduces a reaction-diffusion model of planarian regeneration in which stem cells at wound edges restore head or tail identity by sensing the direction of a wnt-related gradient. If the model is right, it explains how tiny fragments preserve their original head-tail polarity and how regeneration works across a hundredfold range of body sizes.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Sign of the gradient-sensing terms in Eqs. (3.14)-(3.15) is inverted relative to the stated head/tail convention, so the written model would regenerate tail at the anterior wound.","rationale":"The reader correctly identifies the boundary gradient-sensing term as the weakest premise, but the more concrete and serious problem is internal: under the paper's own sign conventions, the written equations send the regeneration signal in the wrong direction. This is not a biological plausibility objection; it is a mathematical check that can be settled directly from Eqs. (3.14)-(3.15) and the initial condition (3.16). If the sign is indeed wrong, the central claim of polarity preservation fails for the model as stated, regardless of whether a molecular mechanism exists. I would not escalate to REJECT because the sign is plausibly a typo and the modeling framework could survive with a corrected sign; however, the conditional acceptance must require a sign-corrected rerun and released code. The reduced-model inconsistency in Eq. (4.5) further supports the need for an explicit correction. The reader's existing conditional verdict already demands code and sensitivity analysis, so the overall disposition remains CONDITIONAL and thus UNCHANGED.","tokens_in":28727,"tokens_out":19788,"duration_ms":208518,"concrete_test":"Run the full model exactly as printed (Eqs. (3.1)-(3.15), Table 1) for the trunk-cut initial condition (3.16) with y0=0.7, α=0.02, and record h and d near both wounds. At x=-L, ∂νw = -w_x<0, so Eq. (3.15) forces Ψ_d>0 while Eq. (3.14) gives Ψ_h=0; the written equations therefore produce d>0 at the anterior wound. Check whether the simulation yields h>0.5 at the anterior wound (polarity preserved) or d>0.5 (polarity reversed). Also test homeostasis with h=0.99 at the left boundary: if Ψ_d>0 persists, the head-left equilibrium is unstable as written.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is polarity-preserving regeneration (§3.3). In the model, head cells h sit at x=-L, tail cells d at x=+L, and w increases from 0 at the head to 1 at the tail (§3.1, Fig. 2.1). For the cutting initial condition (3.16), the residual w-gradient has w_x>0. At the left (anterior) wound, the outward normal derivative is ∂νw = -w_x<0; at the right wound, ∂νw>0. The sensing terms (3.13)-(3.15) are Ψ_h=τ(1-h)(1-d)χ_{>θ}(∂νw) and Ψ_d=τ(1-h)(1-d)χ_{<-θ}(∂νw). Therefore, at the anterior wound Ψ_d is active and Ψ_h is inactive: the model produces tail cells at the edge that should regenerate a head, and head cells at the posterior edge. This is the opposite of the experimental polarity the paper claims to reproduce. The same sign problem destabilizes homeostasis: at the head edge ∂νw<0, so any h<1 triggers tail production. If the simulations in Figs. 3.4-3.7 do preserve polarity, they cannot have been generated by the equations as printed; a sign convention or typo is hidden. Either way, the written model does not support the central claim. The reduced model Eq. (4.5) shows a related inconsistency: its Ψ_c activates both switching terms for the same sign of ∂νw and uses -∂νw in the first term, inconsistent with (3.14)-(3.15).","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces a reaction-diffusion model for planarian regeneration along the anterior-posterior axis. The model tracks stem, head, and tail cell densities together with short-range head/tail signals and a long-range wnt-related signal w. Dynamics in the bulk are supplemented by Wentzell-type dynamic boundary conditions on a boundary compartment, where stem-cell differentiation into head or tail cells is switched by the sign of the normal derivative of w. The authors report numerical simulations of homeostasis, cutting, grafting, and growth, and claim correct reproduction of the main experiments with preservation of polarity over orders of magnitude in body size. The second half of the paper derives reduced two-species and scalar models, analyzes stability of the unpolarized state, and argues that dynamic boundary conditions are necessary for regeneration while Robin-type conditions fail. The paper also predicts oscillations near the threshold of regeneration failure.","tokens_in":29295,"tokens_out":6314,"duration_ms":65693,"significance":"If the central claim is correct, the model offers a minimal, analytically tractable mechanism for polarity-preserving regeneration that does not rely on Turing-type wavelength selection and that scales over large size changes. The explicit model reduction to an order-parameter equation and to a scalar equation, plus the stability analysis in Section 4.4, is a genuine strength: it identifies a concrete positive-feedback loop through boundary sensing and explains why dynamic boundary conditions are qualitatively different from instantaneous Robin conditions. The scalar model's oscillatory instability is a falsifiable prediction. However, the significance is conditional on resolving the sign inconsistency in the sensing terms and on demonstrating that the reported outcomes are robust rather than fine-tuned to near-critical parameter values. As written, the central claim is not supported by the printed equations.","major_comments":[{"comment":"The signs in the gradient-sensing terms are inverted relative to the stated head/tail convention. In the healthy gradient and in the cutting initial condition (3.16), w increases from left to right, so w_x>0. At the left boundary x=-L the outward normal derivative is ∂ν w = -w_x <0, so the indicator χ^ε_{<-θ}(∂ν w) in Eq. (3.15) is active and χ^ε_{>θ}(∂ν w) in Eq. (3.14) is inactive. The written model therefore produces tail cells at the anterior wound and head cells at the posterior wound, which is the opposite of the experimental polarity the paper claims to reproduce. This is not a minor labeling issue: for a trunk fragment with h=d=0, the wrong Ψ term is unsuppressed and drives differentiation in the wrong direction. The reduced model in Eq. (4.5) uses the opposite sign convention (Ψ_c^± contains χ^ε_{>θ}(-∂ν w|±L) for the +1/head state), so the full and reduced systems are internally inconsistent. Either the full model in Section 3 has a sign typo that also affects the reported simulations, or the simulations were generated with a different system than the one printed. In either case, the central demonstration of Section 3.3 is not supported by the equations as written.","section":"Section 3.2, Eqs. (3.13)-(3.15)"},{"comment":"The paper repeatedly claims robust behavior for a wide range of parameter values, but the simulation parameters are explicitly chosen close to critical values. Section 3.3 states: 'Our choices of parameters are close to the critical values, where head and tail regions neither shrink nor expand.' Section 4.3 chooses κ=0.577, which is within 0.001 of the Maxwell point κ*=1/√3≈0.5774 identified in Section 4.2. For a bistable or tristable front system, operating at the Maxwell point makes the size and persistence of grafted head/tail regions highly sensitive to small parameter perturbations. The paper does not provide a systematic parameter scan, a phase diagram, or a basin-of-attraction study. Without such evidence, the qualitative statements about 'robustness' in Section 2.2 and the abstract cannot be distinguished from fine-tuning of the reported simulations.","section":"Sections 3.3 and 4.3"},{"comment":"The abstract and introduction claim that the model 'correctly reproduces' cut and graft experiments and that it is the only reacting-diffusing-species model able to do so. The evidence consists of a small number of simulations with hand-picked initial conditions and no quantitative comparison to experimental data such as regeneration times, minimum fragment sizes, or the observed frequencies of two-headed versus two-tailed outcomes. The paper itself notes in Section 2.2 that the model cannot produce a bias between two-headed and two-tailed animals because head and tail are treated symmetrically. The claims should be softened to 'is consistent with the qualitative outcomes of selected experiments,' and the uniqueness claim should either be removed or supported by a systematic comparison with the models discussed in Section 2.4.","section":"Sections 1, 2.2, and 3.3"}],"minor_comments":[{"comment":"The second sensing function is labeled Ψ_w but should be Ψ_d, consistent with Eq. (3.12) and the surrounding text.","section":"Eq. (3.15)"},{"comment":"The title contains a typo: 'planari an' should be 'planarian.' There are also repeated spacing errors in the text, e.g., 'Differential equations' and 'inhomegeneous Dirichlet boundary conditions' in Section 3.2.","section":"Title and formatting"},{"comment":"The text writes 'critical “Mawell” point' but should read 'Maxwell point.'","section":"Section 4.2"},{"comment":"The numerical simulations are not accompanied by code or data availability statements, and the parameter values in Table 1 are presented without a sensitivity analysis. Providing the simulation code would substantially improve reproducibility, especially given the sign inconsistency noted above.","section":"Section 3.3"},{"comment":"The statement that gradient sensing is 'necessary' to reproduce robust preservation of polarity is presented as a conclusion, but the paper only demonstrates sufficiency of the proposed mechanism. The wording should be adjusted to reflect that this is a conjecture supported by the reduced-model analysis, not a proven necessity.","section":"Section 5"}],"recommendation":"major_revision","confidential_remarks":"The sign inconsistency between the full model and the reduced model is the most serious issue. If the authors confirm that the simulations used the sign convention of Eq. (4.5) rather than Eqs. (3.14)-(3.15), the error may be fixable in revision, and the mathematical core of the paper — the reduction, the stability analysis, and the scalar-model predictions — would remain valuable. I would also ask the editor to require a systematic robustness study before final acceptance, because the near-critical parameter choices currently weaken the paper's central claim. The paper is otherwise within scope for a mathematical biology journal, though the literature claims in the abstract should be moderated."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know up front. The mechanism is genuinely new: dynamic boundary conditions where stem cells at wound edges differentiate into head or tail according to the sign of the normal derivative of a wnt-like long-range signal. That is a real departure from the Turing and French-flag models you usually see in regeneration, and it explains the size robustness in a way those models cannot. The reduction to a scalar order parameter is also well done, and the loss of regeneration under Robin boundary conditions is a genuinely instructive result.\n\nBut there is a load-bearing sign error in the printed equations. At the anterior edge, x=-L, the outward normal derivative of w is negative (w increases towards the tail). The paper's own convention puts head cells at x=-L. Equations (3.14)-(3.15) set Ψ_h to fire on positive ∂νw and Ψ_d (mislabeled Ψ_w) on negative ∂νw. So the written full model regenerates a tail at the anterior wound and a head at the posterior wound. The stress-test note is correct. The reduced model in (4.5) uses the opposite sign in the first boundary term, so the analysis section encodes the intended polarity — but the full model in Section 3, the one whose simulations are shown, does not. Either the simulations were produced by a different sign convention, or the printed equations have a typo. As written, the central claim does not follow from the model.\n\nThe other soft spots are real but less central. The gradient-sensing term is a postulate: no molecular mechanism, no direct measurement, and no independent estimate of the threshold θ or sensitivity ε. Parameters are hand-chosen, and the authors admit they sit close to critical values where head and tail regions neither grow nor shrink. No code or data are supplied, so the simulation-based claims are not independently checkable.\n\nWhat the paper does well deserves credit. The reduction from six species to a scalar equation is careful and honest. The stability calculation for the trivial trunk state shows why dynamic boundary conditions are not just a trick. The oscillation prediction near the recovery-failure transition is concrete and falsifiable. The literature review is thorough and fair.\n\nThis paper is for applied mathematicians and computational developmental biologists. With the sign fixed, code released, and a parameter-sensitivity study added, it would be a useful contribution. In its current form it is not. I would still send it to peer review: a serious referee will catch the sign error and the rest of the analysis is worth the referee's time. But the recommendation should be reject-and-resubmit, not accept.","headline":"The gradient-sensing mechanism is new and the reduction is solid, but the printed boundary terms have the wrong sign — the full model as written would regenerate tail at the anterior wound.","tokens_in":29639,"tokens_out":6948,"would_cite":false,"duration_ms":70185,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["92C15","35K57","35Q92"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper argues that polarity-preserving planarian regeneration is driven by wound-edge stem cells sensing the slope of a wnt-related gradient, not its level, and that a reaction–diffusion model built on this gradient-sensing boundary…","keywords":["planarian regeneration","reaction-diffusion model","dynamic boundary conditions","wnt gradient sensing","polarity preservation","cutting and grafting experiments","tristability","model reduction"],"falsifier":"A concrete test: in an amputated fragment, image a wnt/β-catenin reporter in the first hours after cutting and compare the differentiation fate of cells at the two wounds with the signed slope of the reporter near each wound. If cells at a wound facing a flat or inverted residual gradient still reliably make the correct head or tail, the gradient-sensing mechanism is falsified. A model-level falsifier also exists: the linear stability calculation in Section 4.4 predicts that the unpolarized state is unstable only when the boundary feedback coefficients satisfy $\\tau/\\varepsilon > 1/\\gamma + \\sqrt{2}$ in the scalar reduction; measuring these rates and testing whether regeneration still occurs despite stability would settle the mechanism.","tokens_in":28546,"feed_emoji":"🪱","tokens_out":10935,"duration_ms":107491,"temperature":0.7,"pith_summary":"This paper attempts to show that a compact reaction–diffusion model can reproduce planarian regeneration, with head and tail formed in the correct orientation even from fragments as small as half a percent of the worm. Its key move is to let stem cells at a wound edge read the direction (slope) of a wnt-related signal gradient rather than its absolute concentration, and to implement this as dynamic boundary conditions at the two ends of a one-dimensional body. The authors claim this is the only reacting-and-diffusing-species model that reproduces most cutting and grafting experiments, because it preserves polarity over body sizes spanning two orders of magnitude. If the claim holds, it supplies a minimal biological mechanism: the residual gradient left in a fragment, read at the cut surface, is enough to tell each end which pole to make.","feed_headline":"Wound-edge cells read the wnt slope to fix head and tail in planarians","feed_subtitle":"Sensing gradient direction, not signal level, keeps head and tail correct across cuts, grafts, and 100-fold growth.","key_machinery":"The load-bearing object is the dynamic (Wentzell) boundary compartment: a thin finite-size region at each body edge with its own concentrations and kinetics, coupled to the bulk through diffusive flux and containing an extra wound-healing source that differentiates stem cells according to the sign of $\\partial_\\nu w$ (Eqs. (3.8)–(3.15)). This boundary sensor provides the positive feedback—a small rise in boundary $w$ strengthens the gradient, which strengthens head/tail differentiation, which in turn raises boundary $w$—that makes the unpolarized trunk state unstable and starts regeneration. It also separates the model from Turing-style wavelength selection and from French-flag reading of absolute wnt levels; the reduced analysis distills the action to a scalar equation for $w$ whose stability boundaries predict when regeneration fails.","core_discovery":"On the paper's own terms, the central discovery is that polarity-preserving regeneration can be organized entirely by gradient sensing at body edges, without any cell reading the absolute level of a morphogen. In the model, a long-range wnt-related signal $w$ sits in a monotone gradient from head to tail. After a cut, the remaining fragment keeps a piece of that gradient, and stem cells in a thin boundary compartment at each wound differentiate into head or tail cells according to the sign of the normal derivative $\\partial_\\nu w$, through terms such as $\\tau(1-h)(1-d)\\chi^\\varepsilon_{>\\theta}(\\partial_\\nu w)$ for head and the mirror term for tail. The resulting dynamics regenerate correct poles from tiny trunk fragments and remain robust as the domain grows; the authors also show analytically that replacing the dynamic boundary conditions by instantaneous Robin-type conditions destroys regeneration, because the positive feedback between the boundary value of $w$ and the bulk gradient is lost.","pith_inferences":["A direct experimental test of the paper's mechanism would be to image wnt/β-catenin reporters at a fresh wound and ask whether differentiation direction correlates with the slope of the gradient on the two sides of the cut, rather than with the local concentration.","Because the paper identifies a finite-size boundary compartment as mathematically necessary (dynamic rather than Robin conditions), it implies that a distinct 'pole' tissue with its own kinetics should be experimentally detectable at planarian body edges; failure to find such a compartment would undercut the model.","The conservation of wnt signaling across planarians and hydra suggests the gradient-sensing boundary mechanism could be tested in hydra as well, where the same type of head/foot grafting experiments have been quantified.","If the gradient-sensing rule is really generic, it might be implemented in synthetic tissues or organoids by engineering cells that measure a morphogen slope at an artificial boundary, providing an engineering route to polarity control."],"forward_implications":["A fragment cut from the trunk, head, or tail region regenerates head at the original head-facing edge and tail at the other, with polarity preserved across body sizes from $L=0.005$ to $L=40$ in simulations.","Grafting reproduces the observed either-or outcome: donor head tissue placed near the host head merges and vanishes, while the same tissue placed far from the host head persists and can organize a second axis.","Absolute levels of wnt signaling are not used for polarity in the early regeneration stages, so equally small fragments from different body regions behave alike despite very different local wnt concentrations.","During uniform growth the pattern is maintained for moderate growth speeds; if dilution makes the wnt gradient shallower than the sensing threshold, the head is lost and a second tail replaces it.","Regeneration begins as an instability of the gradientless trunk state, and near the boundary between recovery and failure the model predicts oscillatory wnt concentrations before the gradient collapses."],"supporting_citations":[{"why":"Rink's review supplies the biological basis for stem cells, wound responses, and the 'pole' regions at body edges that motivate the boundary compartment.","marker":"[64]"},{"why":"Supplies the wnt/β-catenin gradient concept and the experimental evidence that graded wnt signaling organizes the AP axis.","marker":"[2]"},{"why":"Shows β-catenin is required for anteroposterior blastema polarity, grounding the choice of wnt as the polarity-carrying signal.","marker":"[52]"},{"why":"Documents a wound-induced wnt expression program that controls regeneration polarity, supporting early asymmetric signaling at wounds.","marker":"[53]"},{"why":"Shows notum, the only polarized early wound gene, inhibits wnt to promote head regeneration; this motivates asymmetry between head and tail wound responses.","marker":"[54]"},{"why":"Documents distinct stem-cell responses to wounds versus tissue absence, motivating strong local differentiation at cut edges.","marker":"[82]"},{"why":"The prior hysteresis-based hydra model that handles grafting experiments; the paper argues it does not robustly produce cutting patterns, providing the comparison baseline for the 'only model' claim.","marker":"[35]"},{"why":"The canonical reaction-diffusion patterning mechanism with domain-size-dependent wavelength; the paper argues it cannot preserve polarity over orders of magnitude.","marker":"[77]"}],"fun_headline_variants":["Wnt slope at wound edge sets head-tail in planarian regeneration","Polarity preserved by sensing wnt gradient sign at wounds","Dynamic wound boundary senses wnt slope, not level, to regrow","Head-tail polarity from gradient direction at cuts, not concentration","Wnt gradient sign at wound edge, not amount, regrows planarians"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that stem cells at a fresh wound can measure the sign (or small thresholded value) of the spatial slope of the wnt-related gradient and turn that into head-versus-tail differentiation; the paper postulates this sensing step rather than deriving it from a molecular mechanism or direct measurement.","fun_headline_variants_meta":{"raw":{"variants":["Wnt slope at wound edge sets head-tail in planarian regeneration","Polarity preserved by sensing wnt gradient sign at wounds","Dynamic wound boundary senses wnt slope, not level, to regrow","Head-tail polarity from gradient direction at cuts, not concentration","Wnt gradient sign at wound edge, not amount, regrows planarians"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000754,"raw_usage":{"total_tokens":3337,"prompt_tokens":913,"completion_tokens":2424,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":529,"completion_tokens_details":{"reasoning_tokens":2333}},"tokens_in":529,"tokens_out":2424,"duration_ms":17718,"temperature":1.0,"reasoning_tokens":2333,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:47:18.833092+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A concrete test: in an amputated fragment, image a wnt/β-catenin reporter in the first hours after cutting and compare the differentiation fate of cells at the two wounds with the signed slope of the reporter near each wound. If cells at a wound facing a flat or inverted residual gradient still reliably make the correct head or tail, the gradient-sensing mechanism is falsified. A model-level falsifier also exists: the linear stability calculation in Section 4.4 predicts that the unpolarized state is unstable only when the boundary feedback coefficients satisfy $\\tau/\\varepsilon > 1/\\gamma + \\sqrt{2}$ in the scalar reduction; measuring these rates and testing whether regeneration still occurs despite stability would settle the mechanism.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Rink's review supplies the biological basis for stem cells, wound responses, and the 'pole' regions at body edges that motivate the boundary compartment."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Shows β-catenin is required for anteroposterior blastema polarity, grounding the choice of wnt as the polarity-carrying signal."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Documents a wound-induced wnt expression program that controls regeneration polarity, supporting early asymmetric signaling at wounds."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Shows notum, the only polarized early wound gene, inhibits wnt to promote head regeneration; this motivates asymmetry between head and tail wound responses."},{"cited_title":"Wenemoser and P","cited_arxiv_id":null,"evidence_quote":"Documents distinct stem-cell responses to wounds versus tissue absence, motivating strong local differentiation at cut edges."},{"cited_title":"Marciniak-Czochra","cited_arxiv_id":null,"evidence_quote":"The prior hysteresis-based hydra model that handles grafting experiments; the paper argues it does not robustly produce cutting patterns, providing the comparison baseline for the 'only model' claim."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The canonical reaction-diffusion patterning mechanism with domain-size-dependent wavelength; the paper argues it cannot preserve polarity over orders of magnitude."}],"review_version":1}