{"id":"35ad8684-d2d4-48d2-a6da-7f6289a2da90","arxiv_id":"2505.08443","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":2.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"A short-range limit of nonlocal aggregation-diffusion equations yields a local thin-film-type model whose four parameters can be tuned to reproduce sorting, engulfment, partial engulfment, and mixing of two cell populations.","lead":"This paper derives a local fourth-order 'thin-film' PDE model of cell aggregation from a nonlocal adhesion model by taking a short-range limit, then uses it to reproduce four observed cell-sorting patterns numerically. A smart generalist might read it as a case study in replacing expensive nonlocal interaction terms with local derivatives while keeping the same qualitative tissue behavior.","discovery_kind":"review","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 'preserves the same phenomenology' claim is asserted, not established: the O(a^4)-truncated local model is explicitly not compared with the nonlocal model, and the numerical evidence only shows hand-picked local parameters reproducing the four Steinberg patterns.","rationale":"The reader identified the Taylor truncation as the weakest assumption, and I agree. The paper is transparent about the heuristic derivation and does not claim mathematical equivalence, but the abstract's 'preserving the same phenomenology' is a central claim that goes beyond the rigorous a→0 convergence. The omission of a direct nonlocal-to-local comparison, combined with the paper's own caveat in §2.1 and the cited findings of [18] for non-compact kernels, makes the claim conditional rather than established. I found no internal inconsistency in the formal derivation that would invalidate the model itself, and the existence theory in [23] provides independent support for the local equation. However, the biological relevance of the simplification depends on finite sensing radii, and that dependence is not quantified anywhere in the manuscript. The proposed side-by-side phase-diagram comparison would settle this concern directly. Since the reader's CONDITIONAL verdict already reflects this gap, no adjustment is needed.","tokens_in":18398,"tokens_out":15091,"duration_ms":144108,"concrete_test":"Run a side-by-side simulation of the nonlocal two-species model (2) and the local model (9) using compactly supported, rescaled potentials φ_ij = K_ij φ with the same K_ij and a finite sensing radius a ∈ {0.05L, 0.1L, 0.2L}. Map K_ij to (κ, α, μ, ω) with the formulas in §3.3, start from the same initial data (e.g., the localized condition in Figure 5), and classify the long-time pattern (sorting, partial engulfment, engulfment, mixing) for a K_12 sweep at fixed K_11 and K_22. If the nonlocal phase boundaries differ from the local model's Figure 4/5 diagram at biologically plausible a, the truncation does not preserve the claimed phenomenology; agreement only in the asymptotic limit a→0 is insufficient.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central assertion is that Eqs. (9) follow from Eqs. (2) in the short-range limit and preserve the same phenomenology. The derivation keeps only the first two terms of the Taylor expansion W_ij*f ≈ -c_ij f - d_ij Δf, dropping the O(a^4 M_4) remainder. Rigorous results [22,34] justify convergence of solutions as a→0, but they do not establish that the truncated local model matches the nonlocal model's metastable dynamics, stationary-pattern selection, or phase boundaries at finite sensing radius a. Section 2.1 explicitly disclaims such a comparison: 'the goal of our paper is not to compare (4) with the nonlocal model (1)'. Section 4 also cites [18] reporting significant local/nonlocal differences for non-compact kernels. The numerical evidence in Figures 4, 5, and 7 shows only that the local model can produce the four Steinberg configurations for selected parameter values; it never compares the two models on the same parameter sweep or initial data. Thus the load-bearing premise that the phenomenology is preserved is asserted rather than tested.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper derives a two-species local aggregation-diffusion system, Eqs. (9), by Taylor-expanding the nonlocal convolution terms in Eqs. (2) to second order, assuming short-range symmetric attractive kernels. It then studies the one-species linear stability and energy dissipation, and uses numerical simulations to argue that the model reproduces the four Steinberg configurations: sorting, partial engulfment, engulfment, and mixing. The paper also surveys recent existence results for the local model and lists open problems.","tokens_in":18655,"tokens_out":10795,"duration_ms":108167,"significance":"If the nonlocal-to-local reduction is faithful at finite sensing radius, the model is a genuinely useful simplification: it has a gradient-flow structure with explicit energy, analytically accessible steady states, four interpretable parameters, cheaper numerical implementation, and a more direct path to calibration than the original nonlocal model. The paper has clear strengths: the formal Taylor derivation and nondimensionalization in Sections 2.1 and 3.1 are internally consistent; the dispersion relation sigma(k) = rho0 |k|^2 (mu2 - |k|^2) recovers the expected Cahn-Hilliard instability condition; the energy-dissipation calculations for F[rho] and F2[rho,eta] are correct under the stated boundary conditions; and the numerical illustrations in Figures 4, 5, and 7, together with the VisualPDE implementations, make the phenomenology accessible. The main unresolved issue is that fidelity to the nonlocal model, which is advertised in the title and abstract, is asserted rather than quantitatively tested.","major_comments":[{"comment":"The abstract's claim that the local model 'preserves the same phenomenology' as the nonlocal model is not established. The Taylor expansion truncates W*f at the second moment and drops an O(a^4 M_4) remainder, and the paper explicitly states that its goal is not to compare Eq. (4) with Eq. (1). The cited rigorous results [22,34] prove convergence in suitable limits as a tends to zero, but they do not quantify the finite-a error in metastable dynamics, stationary-pattern selection, or phase boundaries; the paper itself notes in Section 4 that [18] reports significant local/nonlocal differences for non-compact kernels. Since the central claim of the paper rests on this nonlocal-to-local fidelity, the authors should either provide a quantitative truncation estimate in terms of a and the moments M_4, or include a direct numerical comparison of Eqs. (9) with Eqs. (2) on the same parameter sweeps and initial data over the simulated time scales.","section":"Section 2.1, Eq. (4)"},{"comment":"The claim that the model identifies 'parameter regimes' for sorting, partial engulfment, engulfment, and mixing is supported only by one hand-selected (alpha, omega) point per pattern, with kappa = 2 and mu = 4 fixed. The shaded regions in Figure 4 are not specified by precise inequalities, and no sensitivity analysis or phase-diagram computation is presented. Because the system is a degenerate fourth-order gradient flow and may have multiple metastable stationary states, the manuscript should demonstrate that each pattern occurs on an open parameter set and is robust to initial conditions and numerical resolution before claiming regime identification.","section":"Sections 3.3-3.4, Figures 4, 5, 7"},{"comment":"The parameter relations kappa = K11/K22, alpha = K12/K22, mu = ((M0 - eps/K11)/(M0 - eps/K22)) kappa, and omega = ((M0 - eps/K12)/(M0 - eps/K22)) alpha are derived under the assumption phi_ij = K_ij phi, but the numerical simulations do not state which values of M0, epsilon, K11, K22, and K12 produce the chosen (kappa, mu, alpha, omega). Without this mapping, the connection between the 'clear physical interpretation' of the parameters and the actual simulations remains qualitative; the paper should either make this correspondence explicit or clearly separate the mathematical parameter space from the biological calibration.","section":"Section 3.3, Eq. (13) and surrounding text"}],"minor_comments":[{"comment":"The sentence 'the energy is non-decreasing in time' should read 'non-increasing', since the calculation gives dF/dt <= 0.","section":"Section 2.1"},{"comment":"The finite-volume scheme is only referenced; the paper should state the specific discretization or point to equation numbers in [39], and add at least one grid-refinement check for a two-species steady state to support the numerical claims.","section":"Section 2.3"},{"comment":"The 'weak' and 'strong' cross-adhesion regions should be defined by explicit inequalities rather than only by a sketch, so that the claimed regimes are reproducible by readers.","section":"Section 3.3, Figure 4"},{"comment":"The energy label in the caption appears garbled in the rendered manuscript; please verify the mathematical notation in the final version.","section":"Figure 6 caption"}],"recommendation":"major_revision","confidential_remarks":"The submission appears to build heavily on the authors' own SIAM paper [39], where the local model and its steady states are already developed, and on [23] for existence theory. The editor should verify that the incremental contribution of this manuscript, namely the nonlocal-to-local derivation and the four-pattern numerical study, is sufficiently novel for the target journal, especially since the abstract calls the model 'novel' while the text attributes the model to [39]. In my view, the required technical addition is a quantitative local/nonlocal comparison or truncation-error estimate for the central 'preserving the same phenomenology' claim."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this preprint is a well-written digest of the authors' published local adhesion model [39] and its existence theory [23], not a new result. If you have read those papers, the novel content is thin; if you haven't, this is a good way in. The Taylor-expansion derivation from the nonlocal model is clear, the formal dispersion relation and energy dissipation checks are correct, and the numerics do produce the four Steinberg configurations in one and two dimensions. The paper also earns credit for being honest about scope: it states explicitly that it is not comparing the truncated local model with the nonlocal model, refers to [39] for steady-state and numerical-scheme details, and lists open problems (uniqueness, long-time asymptotics, minimizers on the whole space) rather than pretending they are solved.\n\nThe soft spots are real but proportionate. The abstract's claim that the local model arises as a short-range limit 'while preserving the same phenomenology' is asserted, not demonstrated. The O(a^4 M4) remainder is discarded without a quantitative bound, and the rigorous nonlocal-to-local results [22,34] justify convergence as a goes to zero, not fidelity of the truncated model at the finite sensing radii used in the simulations. Section 2.1 makes this almost explicit with its disclaimer that the goal is not to compare the local and nonlocal models. On the biological side, the pattern map is built from parameters chosen by hand to obtain each configuration; there is no sweep over initial data, no comparison with the nonlocal model, and no comparison with experimental measurements. So the Steinberg patterns are an illustration of model capacity rather than a predictive test. These are limitations, not mathematical errors: the steps shown check out.\n\nI would not cite this preprint in preference to the SIAM paper [39], which contains the model details, steady states, and the numerical scheme. But I would send the manuscript to a referee if the venue is open to synthesis or survey contributions, with the instruction that the 'same phenomenology' claim needs qualification or at least one direct local-versus-nonlocal comparison. For a primary-research journal, the lack of new results is a reasonable desk-reject reason; for a review-oriented journal, it deserves a serious referee.","headline":"A clear, honest synthesis of the authors' earlier local adhesion model, but the 'preserves the same phenomenology' claim is asserted rather than tested.","tokens_in":19193,"tokens_out":2979,"would_cite":false,"duration_ms":32106,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35Q92","92C15","35K55"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper claims that short-range nonlocal adhesion can be replaced by a four-parameter thin-film-like system that still produces the experimentally observed cell-sorting configurations.","keywords":["aggregation-diffusion","cell-cell adhesion","thin-film equation","Cahn-Hilliard equation","differential adhesion hypothesis","cell sorting","nonlocal-to-local limit","gradient flow"],"falsifier":"Run the nonlocal system (2) and the local system (9) with identical compactly supported kernels mapped through the moment formulas, across the parameter plane of Figure 4; any region where the stationary patterns disagree, such as sorting instead of engulfment near the $\\alpha=\\sqrt{\\kappa}$ boundary, or metastable states that persist only in one model, would falsify the claim that the local limit preserves the phenomenology.","tokens_in":18174,"feed_emoji":"🧫","tokens_out":5177,"duration_ms":48548,"temperature":0.7,"pith_summary":"The paper claims that the nonlocal equations commonly used for cell-cell adhesion can, in the short-range limit, be replaced by a system of fourth-order thin-film-like equations with only four parameters. These parameters carry biological meaning: they act as relative surface tensions and adhesion strengths between two cell populations. The authors show numerically, in one and two dimensions, that this local system produces the four configurations predicted by the differential adhesion hypothesis: sorting, partial engulfment, engulfment, and mixing. If the claim holds, the simpler local model gives a tractable route to calibrating adhesion models against experiments and to analytical steady-state predictions that nonlocal models rarely offer.","feed_headline":"Four parameters reproduce all four cell-sorting patterns","feed_subtitle":"Short-range nonlocal adhesion reduces to a thin-film system that sorts, engulfs, or mixes two cell types in simulations.","key_machinery":"The load-bearing object is the reduced local system (9), a system of two fourth-order degenerate parabolic equations of thin-film type. It is obtained by writing each nonlocal convolution as a Taylor expansion in the sensing radius and keeping only the zeroth- and second-moment terms; symmetry kills first-order terms and the error is $O(a^4 M_4)$. The machinery does three jobs: it supplies four interpretable parameters (relative self-adhesion $\\kappa$, cross-adhesion $\\alpha$, and the corresponding population-pressure terms $\\mu$ and $\\omega$), it preserves the gradient-flow structure through the free energy (11), and it permits explicit stationary bump profiles that match numerics.","core_discovery":"On its own terms, the paper establishes that the two-species nonlocal aggregation-diffusion system (2), with convolution kernels $W_{ij}$, formally converges, after Taylor expansion and truncation at second order, to the local system (9): a set of thin-film/Cahn-Hilliard type equations with parameters $\\kappa$, $\\alpha$, $\\mu$, $\\omega$ that are ratios of the first and second moments of the interaction potentials plus the repulsion strength. The central finding is that this local limit preserves the phenomenology of differential adhesion: numerical simulations identify parameter regimes matching the four experimentally observed cell-sorting configurations, with the cross-interaction parameters $\\alpha$ and $\\omega$ controlling the transition from sorting to mixing. The same gradient-flow energy structure of the nonlocal model is retained, giving an energy that decreases along solutions and a basis for existence results reported by the authors in companion work.","pith_inferences":["If the truncation is faithful, the same nonlocal-to-local reduction could be applied to other attraction-repulsion systems, such as crowd dynamics, chemotaxis, or social aggregation, wherever interaction kernels are effectively short-range.","A testable extension would be to map experimental adhesion measurements into the four parameters and predict tissue-sorting outcomes quantitatively, rather than only qualitatively.","The reported metastable coarsening in the weak cross-adhesion regime suggests the local model could serve as a computationally cheap surrogate for studying long-transient dynamics in nonlocal models, a connection the paper only partially explores.","The gradient-flow energy (11) makes this system a convenient testbed for conjectures about Cahn-Hilliard systems with cross-interaction terms, such as wetting-layer behaviour or spinodal decomposition analogies."],"forward_implications":["A nonlocal adhesion model can be simulated and calibrated as a local PDE, avoiding costly convolution evaluations while keeping the qualitative patterns.","The four parameters map directly to measurable or inferable quantities, so inference techniques can target parameters rather than whole interaction kernels.","The local model admits analytical stationary solutions for individual aggregates, something typically unavailable for nonlocal models with general potentials.","Linear stability of the homogeneous state is governed by the sign of $\\mu_2$ ($M_0-\\epsilon$), matching the nonlocal model's condition for existence of stationary states.","Both one- and two-dimensional simulations show the same four experimentally observed configurations, so the simplification is not a one-dimensional artifact."],"supporting_citations":[{"why":"Introduces the continuum nonlocal adhesion model, Eqs. (2), that this paper's local limit starts from.","marker":"[5]"},{"why":"Uses a truncated expansion of the interaction potential to derive a local model, the approach this paper follows and extends to two species.","marker":"[12]"},{"why":"Provides rigorous justification of the nonlocal-to-local limit for two-species systems with compactly supported potentials.","marker":"[22]"},{"why":"Provides rigorous justification of the nonlocal-to-local limit in the one-species case.","marker":"[34]"},{"why":"Establishes existence of weak solutions for the local fourth-order model via its gradient-flow structure.","marker":"[23]"},{"why":"Companion paper supplying the finite-volume numerical scheme, analytical steady-state bump profiles, and the parameter regimes used here.","marker":"[39]"},{"why":"Maps adhesion strengths to the sorting, engulfment, and mixing phase diagram that the simulations reproduce.","marker":"[57]"}],"fun_headline_variants":["Short-range adhesion collapses into thin-film sorting model","Four parameters match all four observed sorting patterns","Local thin-film equations capture differential adhesion sorting","Nonlocal-to-local reduction yields cell sorting patterns"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that truncating the Taylor expansion of the nonlocal convolution after the second moment, valid for small sensing radius and nicely decaying kernels, does not change the biologically relevant dynamics, so that the local model's patterns are the nonlocal model's patterns.","fun_headline_variants_meta":{"raw":{"variants":["Short-range adhesion collapses into thin-film sorting model","Four parameters match all four observed sorting patterns","Local thin-film equations capture differential adhesion sorting","Nonlocal-to-local reduction yields cell sorting patterns"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000734,"raw_usage":{"total_tokens":3254,"prompt_tokens":891,"completion_tokens":2363,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":507,"completion_tokens_details":{"reasoning_tokens":2306}},"tokens_in":507,"tokens_out":2363,"duration_ms":18510,"temperature":1.0,"reasoning_tokens":2306,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T21:55:46.143696+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the nonlocal system (2) and the local system (9) with identical compactly supported kernels mapped through the moment formulas, across the parameter plane of Figure 4; any region where the stationary patterns disagree, such as sorting instead of engulfment near the $\\alpha=\\sqrt{\\kappa}$ boundary, or metastable states that persist only in one model, would falsify the claim that the local limit preserves the phenomenology.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces the continuum nonlocal adhesion model, Eqs. (2), that this paper's local limit starts from."},{"cited_title":"Bernoff and C","cited_arxiv_id":null,"evidence_quote":"Uses a truncated expansion of the interaction potential to derive a local model, the approach this paper follows and extends to two species."},{"cited_title":"Elbar and J","cited_arxiv_id":null,"evidence_quote":"Provides rigorous justification of the nonlocal-to-local limit in the one-species case."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes existence of weak solutions for the local fourth-order model via its gradient-flow structure."},{"cited_title":"Falc´ o, R","cited_arxiv_id":null,"evidence_quote":"Companion paper supplying the finite-volume numerical scheme, analytical steady-state bump profiles, and the parameter regimes used here."}],"review_version":1}