{"id":"ca1c0450-6c1a-4094-9cf4-43ec58465e69","arxiv_id":"2411.09721","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":11,"one_line_summary":"A multi-framework Hes1-Notch model shows that a homogeneous state is unstable, a stable non-homogeneous pattern forms, and this patterning is robust to intrinsic noise.","lead":"This paper builds a spatial model of the Hes1-Notch genetic circuit in brain development, combining ordinary differential equations and a stochastic reaction-diffusion framework. It analyzes when the system switches from oscillations to a stable salt-and-pepper pattern, and checks that the pattern survives cellular noise.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proposition 3.7 does not establish patterned-state stability for the full model: it asserts condition (3.15) controls the non-homogeneous solution, but no equivalence is shown and the degree-10 polynomial is not checked by Routh–Hurwitz.","rationale":"The reader correctly identified parameterization as a limitation, but the more load-bearing issue is internal to the mathematical argument: Proposition 3.7 is the only analytical support for stability of the patterned state in the full ODE model, and its proof is incomplete. This matters because the paper's headline claim combines deterministic patterning with robustness in the stochastic RDME model; if the patterned state were unstable in the full model, the RDME result would lack its deterministic anchor. I do not claim the conclusion is false: the numerics in Figures 3.2, 3.5, and 3.8 are consistent with it, and the characteristic polynomial has special structure that may make the omitted Routh–Hurwitz verification routine. For that reason I would not reject or mark the paper unverdictable; the appropriate disposition is the same conditional acceptance the reader chose, but the condition should explicitly include a spectral check of the patterned steady state and a derivation of the stability boundary for Proposition 3.7. The recommended concrete test is computational but decisive: evaluating the full 10×10 Jacobian at the continued patterned state directly checks the missing equivalence, and a mismatch would falsify the proof as written. The reader's weakest assumption (parameter ad hoc-ness) is real but secondary; it affects biological relevance rather than the internal correctness of the stability argument.","tokens_in":21552,"tokens_out":26072,"duration_ms":246750,"concrete_test":"For the Table 2.1 parameters and the αN-scaling values used in Figure 3.5, continue the non-homogeneous steady state of the two-cell periodic full model (2.2), assemble the 10×10 Jacobian at that state, and compute its eigenvalues. Check whether the maximum real part is negative exactly when the constant term of the Proposition 3.7 polynomial is positive, and compare the sign-change boundary with condition (3.15). If any parameter interval has max Re λ > 0 while p(0) > 0, or the boundary differs from (3.15), Proposition 3.7 is false. If the intervals match, the gap is only in the proof, and a Routh–Hurwitz or Nyquist argument for the degree-10 polynomial would settle the paper's central claim.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The patterned-state leg of the central claim rests on Proposition 3.7, which says the non-homogeneous solution of the full model (2.2) is stable whenever it exists. The proof expands the 10×10 Jacobian, states that all coefficients of the characteristic polynomial are positive except possibly the constant term, and then concludes that 'the stability condition (3.15) again controls the stability.' This does not follow as written. First, the constant term shown in the displayed polynomial is evaluated at the patterned concentrations P1 and P2, whereas condition (3.15) is derived for the homogeneous P0; the proof never shows that p(0)<0 for the patterned state is equivalent to (3.15). Second, for a degree-10 polynomial, positivity of all non-constant coefficients plus a positive constant term is not sufficient for stability: unstable complex conjugate pairs can occur without p(0) changing sign, so the Routh–Hurwitz conditions would need to be verified. Lemma 3.5 only tracks sign changes of p(0) and cannot exclude a complex pair crossing the imaginary axis before p(0) changes sign. The numerical bifurcation diagrams in Figures 3.2 and 3.5 support the intended conclusion, but the analytical proof of full-model patterned stability, which is the basis for the subsequent RDME patterning claim, is incomplete.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops an ODE model of the Hes1-Delta/Notch gene regulatory network on a hexagonal cell grid, with parameters taken from the literature and partly fitted. It derives reduced two-dimensional and scalar models via quasi-steady-state assumptions, analyzes stability of homogeneous and patterned states in a two-cell periodic setting, extends the analysis to the full model, studies patterning on hexagonal tilings, and compares deterministic results with an RDME stochastic model. The main claims are: a unique homogeneous steady state; instability conditions (3.2), (3.15), and (3.19); existence and stability of a non-homogeneous patterned state in the reduced and full models; and preservation of patterning in the RDME model with a patterning coefficient of 0.50 (95% CI 0.43–0.57) at 50 µm³ cell volume.","tokens_in":21884,"tokens_out":6747,"duration_ms":66120,"significance":"If the analytical results were fully established, the paper would provide a useful bridge between deterministic grid ODE models and stochastic RDME models of a well-studied developmental signalling system. Notable strengths are the candid discussion of parameter limitations (Section 4), the reproducible URDME workflow (Section 4.1), and the combination of multiple modelling frameworks. The numerical bifurcation diagrams and RDME simulations give credible qualitative support for the proposed patterning mechanism. However, the central analytical proof of full-model patterned-state stability is incomplete, and the parameters governing the bifurcation are partly fitted to produce the observed behaviours, so the predictive content is currently limited.","major_comments":[{"comment":"The proofs of Propositions 3.6 and 3.7 assert that positivity of all non-constant coefficients of the characteristic polynomial, together with the sign of the constant term, determines stability. This is not valid for polynomials of degree at least three: unstable complex-conjugate root pairs can occur even when all coefficients are positive and the constant term is positive. Consequently, the \"if and only if\" statement in Proposition 3.6 and the conclusion in Proposition 3.7 that \"the stability condition (3.15) again controls the stability\" are not established by the arguments given. In Proposition 3.7 specifically, condition (3.15) is derived for the homogeneous state P0, while the constant term of the characteristic polynomial for the patterned state is evaluated at P1 and P2; no argument connects the sign of that constant term to (3.15). The numerical bifurcation diagrams (Figs. 3.2 and 3.5) support the intended conclusion, but the analytical proof of full-model stability of the patterned state, which underpins the subsequent RDME patterning claim, is incomplete. I recommend either verifying the full Routh–Hurwitz conditions for the specific polynomial structure, or computing the relevant eigenvalues numerically for the reported parameter ranges, and then adjusting the analytical claims accordingly.","section":"§3.2, Propositions 3.6 and 3.7, Eq. (3.15), Lemma 3.5"},{"comment":"The model behaviour is parameterized to produce the desired outputs: the activation rates αi are fitted to PaxDB concentrations and to give the desired dynamics (caption of Table 2.1), KM and Kn are chosen to match overall system behaviour, and k = 1 and h = 4 are selected as the minimum values yielding oscillations. As a result, the instability conditions (3.2), (3.15), and (3.19), and the existence of the patterned state are not independent predictions for the Hes1 system but rather consequences of the chosen parameter values. The paper acknowledges this limitation in Section 4, but the central claim that the patterning is stable under intrinsic noise (Section 3.4) would be considerably strengthened by a sensitivity analysis over KM, Kn, k, and h, which are held fixed without perturbation. This is a scientific rather than mathematical gap; it does not invalidate the analysis conditional on the chosen parameters.","section":"§2.2, Table 2.1, Appendix A, Section 4"},{"comment":"The 95% confidence interval for the patterning coefficient p is obtained by treating high–high couplings as independent Bernoulli trials. Since neighbouring pairs share cells and the spatial pattern is strongly correlated, the effective number of independent samples is smaller than the number of counted couplings, so the reported interval (0.43, 0.57) may be overconfident. A block bootstrap or a spatial resampling procedure would give a more defensible uncertainty estimate. This affects the quantitative statement of patterning coefficient 0.50, though not the qualitative conclusion that patterning is robust to noise for the modelled parameter regime.","section":"§3.4, Fig. 3.8"}],"minor_comments":[{"comment":"The uniqueness of the homogeneous steady state for the full system (2.2) is asserted \"by extension\" without a proof; since the full model involves five species in each cell, a brief argument or a reference would make this step transparent.","section":"§3.1, Proposition 3.1"},{"comment":"The statement that γ′2(¯x1) = γ′2(¯x2) < 1 is justified only \"by inspection\" and by the graphical illustration in Fig. 3.1; an explicit inequality for these derivatives would be more convincing.","section":"§3.1, Eq. (3.7)"},{"comment":"The proof of Proposition 3.9 relies on a \"graphical motivation\" for the double-root condition, and the displayed formula (3.25) would benefit from the intermediate differentiation steps being shown, since the expression is not self-evident.","section":"§3.3, Proposition 3.9, Eq. (3.25)"},{"comment":"The pseudo-species Din is introduced as a diffusing signal but no degradation or decay term is specified; clarify whether Din has a finite lifetime and how the rate αN qkl is derived from the cell-contact signalling process.","section":"§2.3, Eq. (2.11)"},{"comment":"The reproducibility statement refers to the \"Hes1 directory\" in the DLCM workflow but does not provide a DOI, version number, or exact URL for the workflow; adding these would make the results reproducible in practice.","section":"§4.1"},{"comment":"The entries for µD and µN contain garbled typesetting (e.g., \"log 2/50 × 5 log(2) /(45.3, 55.2) × 5\"); the intended expression appears to be 5·log(2)/(45.3–55.2) min⁻¹ and should be typeset accordingly.","section":"Table 2.1"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is within the scope of q-bio.MN and the multi-framework comparison is a genuine strength. The main deficiency is the incomplete stability proof for the full-model patterned state, which is load-bearing for the paper's central claim. Given the sparsity of the 10×10 Jacobian, a numerical eigenvalue verification over the reported parameter range would be straightforward and would likely settle the issue. The parameterization concerns are secondary but should be addressed with sensitivity analysis or a more guarded interpretation of the predictions."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"We've got a Hes1-Notch modeling paper that tries to bridge three frameworks: a 5-variable ODE grid model, reduced 2D and scalar ODEs, and an RDME stochastic model. The new material is the specific 5-variable model and the attempt to rigorously connect its linear stability to the reduced systems, plus a three-cell pattern-selection analysis and a stochastic version with a diffusing Dll1 pseudo-species. The reduced systems (2.3)-(2.4) and the scalar limit (2.5) are derived carefully, and the two-cell stability analysis of the reduced model is complete and correct as far as I can see. Proposition 3.2 is simple and valid, and Proposition 3.4 works because the relevant polynomial is quadratic. The comparison to Collier's model is honest and useful.\n\nThe soft spots are where the paper extends beyond the reduced model. Proposition 3.7 is the main one: the characteristic polynomial for the 10x10 Jacobian around the patterned state is degree 10, and 'all coefficients positive except possibly the constant term' does not imply stability. You need Routh-Hurwitz, or at least an argument that no complex pair crosses the imaginary axis before the constant term changes sign. The proof says condition (3.15) 'controls stability,' but (3.15) was derived for the homogeneous state, and the constant term of the patterned-state polynomial is not shown to be equivalent to it. Lemma 3.5 only tracks sign changes of the zero-order term; it doesn't rule out complex instability. The numerical bifurcation diagrams suggest the conclusion is true, but the analytic proof as written is incomplete. Propositions 3.3 and 3.6 have similar 'by extension' and 'by inspection' jumps, though these are less serious because the reduced-model stability analysis is solid.\n\nThe parameterization is another soft spot: the α_i are fitted to PaxDB concentrations and to produce oscillations; K_M and K_n are chosen to match overall behavior; k=1 and h=4 are minimal values that give oscillations. The authors acknowledge this in the Discussion and provide confidence intervals from a Monte Carlo perturbation, which is good practice, but it means the quantitative predictions are conditional on choices that aren't independently constrained. The quasi-steady-state reductions assume N, P, n are fast, and that's not validated against the full model dynamics; the claim that the reduced model captures steady-state behavior is fine, but the transient comparison in Fig. 2.3 is only qualitative.\n\nOverall, this is a useful modeling paper for anyone working on Hes1 or Delta-Notch patterning. The reduced-model analysis is a legitimate contribution, and the RDME comparison is a nice example of linking deterministic and stochastic frameworks. The gaps in the full-model proofs are fixable and don't undermine the central qualitative claim, which is supported numerically. I'd send it to peer review with a request for a rigorous revision of Proposition 3.7 and an explicit statement about parameter identifiability. I would cite it if I were working on this system.","headline":"A genuinely useful multi-framework Hes1-Notch modeling paper whose reduced-model stability analysis is sound, but the full-model patterned-stability proof has a real gap and the parameterization is partly fitted.","tokens_in":22425,"tokens_out":4009,"would_cite":true,"duration_ms":38501,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["92-10","92B25","92C15","34A33","60J20","34C60","34F10"],"pacs":[],"model":"deepseek-v4-flash","headline":"A minimal spatial model of the Hes1-Notch network produces transient oscillations followed by a stable salt-and-pepper cell-fate pattern that survives intrinsic noise.","keywords":["Hes1","Notch signaling","lateral inhibition","pattern formation","genetic oscillator","quasi-steady-state reduction","reaction-diffusion master equation","neural progenitor differentiation"],"falsifier":"Measure the model's parameters (activation rates, half-lives, Hill constants, and cell-cell coupling strengths) from mouse neural progenitor cells, insert them into condition (3.15), and check whether the left-hand side is below $-1$; if it is not, the predicted loss of homogeneity does not occur in the model with measured parameters.","tokens_in":21344,"feed_emoji":"🧬","tokens_out":7618,"duration_ms":91220,"temperature":0.7,"pith_summary":"Hes1 is a transcription factor whose levels oscillate in neural progenitors, and the final high/low pattern of Hes1 expression is tied to whether a cell becomes a neuron or a glial cell. This paper tries to establish that a minimal spatial model of the Hes1-Notch network—five molecular species per cell on a hexagonal grid—can reproduce both the damped oscillations and the final salt-and-pepper pattern without adding external regulation. The authors reduce the five-species equations to two-dimensional and scalar systems by assuming some molecules reach steady state quickly, then prove that the homogeneous state is unstable under explicit conditions and that a non-homogeneous patterned state exists and is stable. They also show in a stochastic reaction-diffusion master equation version that the pattern survives intrinsic noise, with a patterning coefficient of 0.50 (95% CI 0.43–0.57) at a physiological cell volume of 50 µm3. If this is right, the oscillatory-then-sustained expression modes seen in development may be intrinsic to the Hes1-Notch gene regulatory network itself.","feed_headline":"Stable salt-and-pepper cell fates arise from a Hes1-Notch model","feed_subtitle":"Even with intrinsic noise, model keeps near-perfect high/low Hes1 patterning at cell volumes seen in embryos.","key_machinery":"The object that carries the argument is a pair of decreasing Hill-type functions, $f(x)=1/(a+x^k)$ and $g(x)=1/(1+b x^h)$, which arise after quasi-steady-state reductions and represent Hes1-mediated repression and Notch-mediated activation of Hes1 mRNA. The stability analysis turns on the derivative balance $f(\\bar{x}_0)g'(\\bar{x}_0)-f'(\\bar{x}_0)g(\\bar{x}_0)<-1$: it decides whether a perturbation from the homogeneous state grows or decays, and the same balance appears in the full five-species model as condition (3.15) through a determinant-preservation lemma for the state reduction. For the spatial pattern, Fourier analysis on the hexagonal lattice selects the mode with period three lattice steps, and a three-cell model with coupling matrix $W_3$ serves as the minimal system that reproduces the full grid's bifurcation structure.","core_discovery":"The paper's central claim is that the Hes1-Notch signalling network, written as a five-variable ODE per cell coupled over a regular hexagonal grid, settles from transient oscillations into a stable static pattern in which neighbouring cells express high and low Hes1 alternately. Using quasi-steady-state reductions, the paper derives reduced equations of the form $\\dot{x}=\\langle y_{\\mathrm{in}}\\rangle f(x)-x$, $\\dot{y}=v(g(x)-y)$ and a scalar form $\\dot{x}=\\langle g(x_{\\mathrm{in}})\\rangle f(x)-x$, where $f$ and $g$ are decreasing Hill-type functions. Propositions 3.2, 3.3, 3.4, 3.6, and 3.7 show that the homogeneous steady state is unstable exactly when $f(\\bar{x}_0)g'(\\bar{x}_0)-f'(\\bar{x}_0)g(\\bar{x}_0)<-1$ (equivalently condition (3.15) in the full model), and that whenever a non-homogeneous solution exists it is stable. On a hexagonal lattice the most unstable Fourier mode has period three lattice steps, matching the observed checkerboard pattern, and a three-cell reduction with coupling matrix $W_3$ reproduces the bifurcation behaviour. In the stochastic RDME formulation, the patterned state remains close to perfect even at high intrinsic noise, with patterning coefficient $\\hat{p}=0.50$ (95% CI 0.43–0.57) at 50 µm3.","pith_inferences":["A direct test of the model's core mechanism would be to measure the patterning coefficient in living neural progenitors with lineage tracing; if the observed high-high coupling is not near 1/2, the specific salt-and-pepper prediction would fail even if the instability condition holds.","The same quasi-steady-state reduction strategy could transfer to other lateral-inhibition systems, such as Notch-Delta models with different Hill exponents, to predict whether they also exhibit a three-cell minimum pattern.","Because the parameters $k=1$ and $h=4$ were chosen as the smallest values producing oscillations, the framework suggests a parameter scan of steeper Hill coefficients (or added delay) could make oscillations less damped without changing the patterning result.","If real neural progenitors show noise-robust patterning, the RDME result implies that stochasticity acts on the timing of fate decision rather than on the pattern itself, a distinction that single-cell tracking could resolve."],"forward_implications":["The explicit instability conditions (3.2), (3.15), and (3.19) give parameter checks for predicting when a uniform Hes1 population will develop into a high/low patterned state.","The same reduction framework, with identical parameters $a$ and $b$ but different time-scale parameter $v$, yields seven reduced systems, so the timing of fate decision, not the steady-state pattern, is what changes between reductions.","The stochastic RDME model preserves checkerboard patterning at cell volumes down to 1 µm3, and at 50 µm3 the patterning coefficient is 0.50 (95% CI 0.43–0.57), close to perfect alternation.","Damped oscillations followed by stable high/low expression can arise from the Hes1-Notch network alone, suggesting the two expression modes need not be imposed by external signals.","If Notch activation strength is weakened, the model predicts the patterned state is lost through a bifurcation, which could be tested by titrating Notch signalling in a neural progenitor population."],"supporting_citations":[{"why":"Supplies the lateral-inhibition pattern-formation model and stability-analysis template that the reduced equations extend.","marker":"[10]"},{"why":"Provides the experimental evidence for the Hes1 negative feedback loop and the half-life values used for mRNA and protein degradation.","marker":"[21]"},{"why":"Provides experimental observations of oscillatory dynamics and salt-and-pepper patterning in mouse neural progenitors.","marker":"[26]"},{"why":"Describes the Hes1 regulatory interactions and the neuron/glial fate decision used to draw the network schematic.","marker":"[45]"},{"why":"Motivates the full ODE system as a Notch-Hes signalling model with multistability and pattern-transition behaviour.","marker":"[38]"},{"why":"Supplies protein abundance ratios used to fit the activation rates $\\alpha_i$ to biological concentrations.","marker":"[24]"},{"why":"Supports the conclusion that stochastic behaviour in the Hes1 pathway is stable to noise, a result the RDME model is compared with.","marker":"[39]"}],"fun_headline_variants":["Hes1 model yields stable salt-and-pepper cell fates","Noise-resistant pattern from Hes1-Notch equations","Checkerboard fates emerge despite intrinsic noise","Math explains stable fate choice in neural development","Neural pattern proven stable even with stochastic noise"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The parameters controlling activation, repression, and Hill coefficients are fitted or chosen ad hoc—not directly measured—so the predicted instability and patterning may not hold for real Hes1 systems if those values are wrong.","fun_headline_variants_meta":{"raw":{"variants":["Hes1 model yields stable salt-and-pepper cell fates","Noise-resistant pattern from Hes1-Notch equations","Checkerboard fates emerge despite intrinsic noise","Math explains stable fate choice in neural development","Neural pattern proven stable even with stochastic noise"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000222,"raw_usage":{"total_tokens":1508,"prompt_tokens":1051,"completion_tokens":457,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":667,"completion_tokens_details":{"reasoning_tokens":382}},"tokens_in":667,"tokens_out":457,"duration_ms":5425,"temperature":1.0,"reasoning_tokens":382,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T20:43:00.604020+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the model's parameters (activation rates, half-lives, Hill constants, and cell-cell coupling strengths) from mouse neural progenitor cells, insert them into condition (3.15), and check whether the left-hand side is below $-1$; if it is not, the predicted loss of homogeneity does not occur in the model with measured parameters.","supporting_citations":[{"cited_title":"Pattern Formation by Lateral Inhibition with Feedback: a Math- ematical Model of Delta-Notch Intercellular Signalling","cited_arxiv_id":null,"evidence_quote":"Supplies the lateral-inhibition pattern-formation model and stability-analysis template that the reduced equations extend."},{"cited_title":"Oscillatory expression of the BHLH factor Hes1 regulated by a negative feedback loop","cited_arxiv_id":null,"evidence_quote":"Provides the experimental evidence for the Hes1 negative feedback loop and the half-life values used for mRNA and protein degradation."},{"cited_title":"Oscillatory control of factors determining multipotency and fate in mouse neural progenitors","cited_arxiv_id":null,"evidence_quote":"Provides experimental observations of oscillatory dynamics and salt-and-pepper patterning in mouse neural progenitors."},{"cited_title":"Multistability and transitions between spatiotemporal patterns through versatile Notch-Hes signaling","cited_arxiv_id":null,"evidence_quote":"Motivates the full ODE system as a Notch-Hes signalling model with multistability and pattern-transition behaviour."},{"cited_title":"PaxDb 5.0: Curated Protein Quantification Data Suggests Adaptive Proteome Changes in Yeasts","cited_arxiv_id":null,"evidence_quote":"Supplies protein abundance ratios used to fit the activation rates $\\alpha_i$ to biological concentrations."},{"cited_title":"Stochasticity in the miR-9/Hes1 oscillatory network can account for clonal heterogeneity in the timing of differentiation","cited_arxiv_id":null,"evidence_quote":"Supports the conclusion that stochastic behaviour in the Hes1 pathway is stable to noise, a result the RDME model is compared with."}],"review_version":1}