{"id":"022c6dd3-1d9f-4093-a3ae-7fecb7077733","arxiv_id":"2505.23988","paper_version":1,"verdict":"REJECT","confidence":"HIGH","novelty_score":4.0,"correctness_risk":"high","formal_verification":"none","parameter_count":8,"one_line_summary":"Heat equation solutions are transformed with Richards growth functions and fitted to scratch assay cell density data; the fits recover m near 1, but the reaction-diffusion connection is unproven.","lead":"The paper builds cell-density models by feeding heat equation solutions through Richards growth functions, then fits the curves to scratch assay data. The fits look reasonable, with Richards exponent m near 1, but the claimed reaction-diffusion basis is not actually derived.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Ansatz I has no growth term: differentiating Eq. (6) yields a nonlinear diffusion equation, not a reaction-diffusion equation, so the central proliferation claim is unsupported for half the reported fits.","rationale":"The reader's weakest_assumption flags the absence of a derived PDE for u. Direct differentiation confirms and sharpens this: Ansatz I yields no growth source, so the problem is not merely an omitted derivation but a mathematical contradiction with the modeling claim. This is load-bearing because the title, abstract, and conclusion claim proliferation and diffusion; if the ansatz used for half the fits is purely diffusive, the estimated parameters (m,b) are not Richards growth parameters in the reaction-diffusion sense. Ansatz II does contain a source, so part of the paper's positive evidence may survive, but the paper does not state that PDE or justify the extra gradient term biologically, and it treats the two ansatze as equivalent. The paper provides no machine-checked proof or reproducible code to offset this gap. Since the reader already rejected the paper and this analysis reinforces that rejection on a more concrete basis, the verdict remains unchanged.","tokens_in":13668,"tokens_out":7228,"duration_ms":71240,"concrete_test":"Take h(x,t)=t+x^2/2, which satisfies the heat equation h_t=h_xx=1. For the Richards F(u), define u by inverting Eq. (6) under Ansatz I. Evaluate u_t-u_xx and F(u) at (x,t)=(1,1). The equality u_t-u_xx=F(u) will fail; instead u_t-u_xx=-(F'(u)/F(u))u_x^2. If this holds, Ansatz I is not a reaction-diffusion proliferation model, and the paper must either derive the correct PDE or remove Ansatz I from the central claim.","verdict_should_be":"UNCHANGED","load_bearing_attack":"To support 'cell proliferation and diffusion', the constructed u(x,t) must satisfy a reaction-diffusion equation with a growth source. For Ansatz I (Methods, Eqs. (4)-(6); Table II), this fails. With u=g(h) and dg/dh=F(g), h_x=u_x/F(u), h_t=u_t/F(u), and h_xx=u_xx/F(u)-F'(u)u_x^2/F(u)^2. Since h_t=h_xx, u_t=u_xx-[F'(u)/F(u)]u_x^2. No F(u) term appears; the Richards function contributes only to the nonlinear diffusion coefficient. For Richards F, h=t gives the ODE because u_x=0, but a genuine space-time heat solution does not. Thus Ansatz I fits shown in Figs. 2-4 are nonlinear-diffusion curve fits, not evidence of proliferation plus diffusion. Ansatz II (Eqs. (7)-(8)) does yield u_t=u_xx+F(u)+[(1-F'(u))/F(u)]u_x^2, containing a source, but this PDE is never stated and the paper presents the two ansatze as interchangeable. The derivation is deferred to a missing supplement; direct differentiation settles it.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes two nonlinear transformations of heat-equation solutions (Ansatz I and Ansatz II) driven by a Richards growth ODE, claims that the resulting density fields u(x,t) model cell proliferation and diffusion, and fits the parameters m and b to scratch-assay data from three melanoma cell lines (12505Lu, WM983C, C8161). The heat-equation input h(x,t) is approximated by a fourth-order two-variable Hermite polynomial expansion. Parameter estimation is done first by least squares / log-likelihood and then with an additive fractional Brownian motion field. The authors report optimal m and b values, claim agreement with previous growth-rate estimates, and state that the solutions describe the spatiotemporal patterns of the experimental cell densities.","tokens_in":13920,"tokens_out":5798,"duration_ms":54544,"significance":"If the construction were valid, the paper would provide closed-form analytic solutions for reaction-diffusion models with a Richards nonlinearity, which could be useful in cell-biology modeling. The authors make a serious attempt to compare their formulas against real experimental data and to add a noise-robustness analysis. However, the central mathematical claim is not established in the manuscript: the PDE satisfied by u(x,t) is never stated, Ansatz I in fact does not yield a reaction term, and several inconsistencies in the parameter-search ranges and time handling make the reported estimates irreproducible. The in-sample validation does not provide independent confirmation of the model.","major_comments":[{"comment":"Ansatz I does not produce a reaction-diffusion equation with a growth term. Direct differentiation of u = g(h) with dg/dh = F(u) and h_t = h_xx yields u_t = u_xx - [F'(u)/F(u)] u_x^2, which is a nonlinear-diffusion equation with no source term F(u). Thus the fits shown for Ansatz I in Figs. 2-4 do not support the paper's central claim of modeling 'cell proliferation and diffusion'; they are curve fits of a density-dependent diffusion process. The derivation is said to be in the supplementary material, but the claim requires a stated PDE in the main text.","section":"§Methods, Eqs. (4)-(6) and Table II"},{"comment":"The PDE satisfied by u(x,t) is never stated for either ansatz. The paper defines the transformation G(u)=g(f(t)h), applies the heat operator formally, and then jumps to the integral relations (6) and (8) and the closed forms in Table II. Without the resulting u_t = ... equation, the interpretation of these formulas as solutions of a reaction-diffusion problem is unverifiable. The derivation is deferred to a supplementary file that is not included in the preprint and cannot be checked.","section":"§Methods, Eqs. (3)-(8)"},{"comment":"The contour plots of the error and log-likelihood surfaces vary m only from 1 to 2, yet the reported optimal values in Eqs. (20)-(21) are m = 0.5, 0.639, 0.54, 0.75, all below 1. The minima shown in the figures are therefore not the reported optima, and the text's statement that 'm values near 1' capture the behavior is inconsistent with the reported estimates. The figures cannot substantiate the claimed 'distinct minimum' of the error surfaces for the parameters finally used.","section":"§Results, Figs. 2-4 and Eqs. (20)-(21)"},{"comment":"The time variable is never reconciled with the experimental times. The data are collected at t = 0, 16, 32, 48 h (or 0, 6, 12, 18 h for C8161), while the Ansatz II solution involves e^t and is plotted only for t ∈ [0,2]. No non-dimensionalization of time is given, and a=1 is fixed without explanation. With t in hours, the term (e^t h)^{-a/m} in Table II would cause the density to saturate almost immediately, contradicting the observed gradual filling of the scratch. The parameter estimates are therefore not reproducible without a clear time-scaling convention.","section":"§Results, Table II and Eqs. (20)-(21)"},{"comment":"The claimed confirmation that the solutions 'well describe' the experimental patterns is based on the same least-squares or log-likelihood fit that produced the parameters. The fBM robustness step adds noise and re-estimates the same parameters on the same data, so it does not break the circularity. An independent validation—e.g., held-out time points, model-selection criteria such as AIC (which the Discussion mentions as future work), or comparison against a null diffusion-only model—is required before the descriptive claim can be made.","section":"§Parameter Estimation, Eqs. (14)-(19) and §Discussion"}],"minor_comments":[{"comment":"The initial condition expression is missing parentheses: the intended form appears to be u(x,0) = 1 / (η + α/(1-α) e^{-x^2/(2β^2)}). The current printed formula is ambiguous.","section":"§Data, Eq. (11)"},{"comment":"There is a typo: 'otsude' should be 'outside'.","section":"§Data, text near Eq. (11)"},{"comment":"The symbol m is used both for the Hermite truncation order (Eq. 13) and for the Richards nonlinearity exponent throughout the results; this notational collision is confusing and should be resolved.","section":"§Solutions using Hermite polynomials, Eq. (13)"},{"comment":"The justification for truncating the Hermite expansion at order 4 is only 'the higher order terms are O(1/n!)'; this is not a rigorous uniform error bound for the specific initial condition in Eq. (11). A convergence test or an a posteriori error estimate should be reported.","section":"§Discussion, limitations paragraph"}],"recommendation":"reject","confidential_remarks":"The manuscript repeatedly refers to a supplementary file for the core derivation of the PDE and the integral relations; that file is not available with the preprint, which makes the main result impossible to verify. The inconsistency between the plotted m∈[1,2] search range and the reported optima m<1 suggests that the parameter estimation displayed in the figures was not the one used to produce the reported values. These issues go beyond presentation and affect the validity of the central claims."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know two things before reading this paper. First, the explicit constructions in Table II — Richards-transformed heat solutions written with two-variable Hermite polynomials — are new, and applying them to scratch assay data is a reasonable exercise. Second, and more importantly, the paper's central claim that both ansatze model proliferation plus diffusion is only half true. For Ansatz I, differentiating ∫du/F(u)=h gives u_t = u_xx − (F'/F) u_x^2. There is no F(u) term. That is a density-dependent diffusion equation, not a reaction-diffusion equation. So when the paper fits Ansatz I to the scratch assay data (Figs. 2–4, top rows), it is curve-fitting with a nonlinear diffusion model, not validating a growth-plus-diffusion mechanism. Ansatz II does yield a source term: u_t = u_xx + [(1−F')/F] u_x^2 + F(u). The paper never states this PDE, and it presents the two ansatze as interchangeable, which they are not.\n\nThe paper does some things well. The parameter estimation pipeline is straightforward, and the authors are transparent about two limitations: the Hermite truncation is not rigorously justified, and the initial conditions are fixed. They also admit the derivation is in a missing supplement. But the inconsistencies are hard to ignore. The contour plots search m from 1 to 2, yet the reported optima include m=0.5, 0.54, 0.639. The parameter a is set to 1 with no sensitivity analysis. The 'confirmation' that the solutions describe the data is the same least-squares fit that produced the parameters, and the fBM robustness step simply adds a noise model on top of that same fit. None of this is fatal by itself, but together it means the evidence for the model is much weaker than the abstract claims.\n\nThe load-bearing flaw is the Ansatz I misinterpretation. If a referee is willing to ignore that, they might accept the paper as a phenomenological fitting exercise. But the title and abstract claim proliferation and diffusion, and that claim is unsupported for half the reported fits. The paper deserves a serious referee because Ansatz II is formally correct and the application is useful, but it needs major revision: either drop Ansatz I, reinterpret it as nonlinear diffusion, or provide the missing derivation and show that the fitted results still make sense. I would not cite this in its current form, but I'd bring it to a reading group as a cautionary example of how easy it is to misread a Hopf-Cole-style transformation.","headline":"Ansatz I is a nonlinear diffusion equation, not a reaction-diffusion equation, so half the fits don't actually model proliferation; the paper needs major revision but has a salvageable core in Ansatz II.","tokens_in":14476,"tokens_out":3181,"would_cite":false,"duration_ms":28840,"reading_group":"maybe","serious_thinker":"no","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper claims that nonlinear transformations of heat equation solutions yield analytic cell-density fields that match scratch assay data across three cell lines.","keywords":["Hopf-Cole transform","Richards growth function","cell proliferation","scratch assay","Hermite polynomials","fractional Brownian motion","reaction-diffusion","parameter estimation"],"falsifier":"Numerically solve the intended reaction-diffusion equation $u_t = u_{xx} + F(u)$ with the fitted Richards parameters and the same initial condition, then compare those profiles with the Table II solutions at the same time points; if they diverge appreciably, the transform solutions are not the reaction-diffusion solutions and the claim of describing proliferation plus diffusion collapses.","tokens_in":13418,"feed_emoji":"🧫","tokens_out":3976,"duration_ms":36298,"temperature":0.7,"pith_summary":"This paper attempts to show that the spatial and temporal cell-density patterns seen in scratch assays can be reproduced by analytic formulas built from nonlinear transformations of heat equation solutions. It constructs two such transformations using the Richards growth function, represents the heat-equation ingredient with two-variable Hermite polynomials, and fits the resulting expressions to experimental data from three cell lines. The fitted parameters are robust to fractional Brownian noise and agree with previous estimates. If the claim holds, the approach offers a computationally cheap analytic alternative to numerical reaction-diffusion simulations for cell proliferation and migration.","feed_headline":"A heat-equation shortcut reproduces scratch-assay cell patterns","feed_subtitle":"Analytic solutions built from Richards growth and Hermite polynomials fit three cell lines with robust parameters.","key_machinery":"The engine of the paper is the Hopf-Cole-style nonlinear invertible transformation: a change of variables that maps solutions of the linear heat equation into candidate solutions of a nonlinear growth-diffusion process via cancellation of terms. Two ansatze operationalize this idea for the Richards growth ODE, yielding the explicit formulas in Table II; two-variable Hermite polynomials provide a series representation of the heat solution $h(x,t)$ from the scratch-assay initial condition; and fractional Brownian motion fields supply a noise model for robust parameter estimation.","core_discovery":"The central discovery is that a cell-density field $u(x,t)$ can be generated by composing a solution $h(x,t)$ of the linear heat equation with a nonlinear growth map, in the spirit of the Hopf-Cole transformation. Two ansatze are derived: Ansatz I sets $u=g(h)$ with $dg/dh=F(g)$, and Ansatz II sets $G(u)=e^t h$ with $\\frac{d}{du}\\log G(u)=1/F(u)$. For the Richards growth function $F(u)=a u(1-(u/b)^{1/m})$, these produce explicit closed forms (Table II). Using a fourth-order Hermite-polynomial representation of the scratch initial condition, these formulas fit the measured density profiles of the 12505Lu, WM983C, and C8161 cell lines at four time points each, with fitted $m$ near 1 and $b$ around 0.15--0.25, and the parameters remain stable when fractional Brownian noise is added to the residuals.","pith_inferences":["A direct test would be to derive explicitly the partial differential equation that the transformed density $u(x,t)$ satisfies; if that PDE contains no genuine growth term, the fits are curve fitting rather than reaction-diffusion modeling.","The method could be used predictively by fitting parameters to early time points and then forecasting later density profiles, a task not demonstrated in the paper.","The same transform recipe might generalize to advection-diffusion or other linear PDEs to model biased cell migration, which the paper does not explore.","Because the initial condition is fixed from the $t=0$ data, letting the initial-condition parameters vary in the fitting procedure could either sharpen or break the claimed agreement."],"forward_implications":["If correct, the analytic formulas in Table II can replace numerical solution of reaction-diffusion PDEs when modeling scratch-assay cell dynamics.","Recovered values of $m$ near 1 support logistic-like growth for the tested cell lines, consistent with prior biological observations.","Robust parameter estimates under fractional Brownian noise suggest the fitting procedure is usable on noisy experimental data.","The two ansatze provide alternative temporal behaviors (no modulating factor versus $f(t)=e^t$), which can be matched to different data regimes.","The transformation framework extends to other growth ODEs by substituting a different $F(u)$ into the same ansatze."],"supporting_citations":[{"why":"Supplies the scratch-assay experimental data for the three cell lines at multiple time points that the model is fitted to.","marker":"[30]"},{"why":"Provide the Hopf-Cole transformation idea that underlies the nonlinear mapping from heat solutions to density fields.","marker":"[17,18]"},{"why":"Define the Richards growth function used as the nonlinearity $F(u)$ in both ansatze.","marker":"[33,34]"},{"why":"Give the two-variable Hermite polynomials used to represent the heat-equation solution from the initial condition.","marker":"[36]"},{"why":"Supplies previous parameter estimates for $m$ that the paper's fitted values are compared against.","marker":"[32]"},{"why":"Defines the fractional Brownian motion field used to model residual noise in the robustness analysis.","marker":"[37]"},{"why":"Connects go-or-grow cell-cycle models with density-dependent diffusion, motivating the class of solutions considered.","marker":"[15]"},{"why":"Provides the log-likelihood and model-selection framework used for parameter estimation and comparison.","marker":"[14]"}],"fun_headline_variants":["Heat equation trick models cell spread","Nonlinear heat solutions fit cell data","Hermite-Richards model reproduces assay patterns","Cell migration captured by heat transforms","Scratch assay patterns from heat equation"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the transformed density $u(x,t)$ actually represents a cell population that both grows and diffuses; the paper never derives the partial differential equation that $u$ satisfies, so the proliferation interpretation rests on an assumed Hopf-Cole-style cancellation rather than on a derived growth term.","fun_headline_variants_meta":{"raw":{"variants":["Heat equation trick models cell spread","Nonlinear heat solutions fit cell data","Hermite-Richards model reproduces assay patterns","Cell migration captured by heat transforms","Scratch assay patterns from heat equation"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000343,"raw_usage":{"total_tokens":1857,"prompt_tokens":889,"completion_tokens":968,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":505,"completion_tokens_details":{"reasoning_tokens":905}},"tokens_in":505,"tokens_out":968,"duration_ms":9625,"temperature":1.0,"reasoning_tokens":905,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T12:38:32.469794+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Numerically solve the intended reaction-diffusion equation $u_t = u_{xx} + F(u)$ with the fitted Richards parameters and the same initial condition, then compare those profiles with the Table II solutions at the same time points; if they diverge appreciably, the transform solutions are not the reaction-diffusion solutions and the claim of describing proliferation plus diffusion collapses.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the scratch-assay experimental data for the three cell lines at multiple time points that the model is fitted to."},{"cited_title":"A., Mansfield, E","cited_arxiv_id":null,"evidence_quote":"Give the two-variable Hermite polynomials used to represent the heat-equation solution from the initial condition."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies previous parameter estimates for $m$ that the paper's fitted values are compared against."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the fractional Brownian motion field used to model residual noise in the robustness analysis."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the log-likelihood and model-selection framework used for parameter estimation and comparison."}],"review_version":1}