{"id":"242c8ddc-cf5f-4a36-a63b-3b13af1715b3","arxiv_id":"1908.07048","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A delay integro-differential equation models heterogeneous, plastic stem cell populations in cancer, but the framework and applications mostly restate the author's prior work.","lead":"This paper lays out a mathematical framework that describes how cancer cell populations evolve when individual cells carry different epigenetic states and can switch states during division. It also states four open mathematical problems that would have to be solved before the framework can become a truly predictive clinical tool.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The transition kernel p(x,y) is unmeasured and its Beta form is justified only by the author's own simulations, so the hybrid-model applications in Sections 4.2-4.3 are conditional on an unvalidated assumption.","rationale":"The reader's weakest assumption identifies the same load-bearing point: the inheritance kernel p(x,y) is unmeasured, and the Beta form is asserted from the author's own simulations rather than independent experimental data. I agree with that assessment. The derivation of Eq. (11) itself is defensible as a mathematical framework for arbitrary p, so the framework claim is not fatally flawed. However, the abstract and strongest claim also assert that hybrid implementations reproduce two biological scenarios, and those applications inherit the unvalidated kernel. Because the paper is explicitly a framework/exposition paper and flags the inverse problem of estimating p as open, the honest verdict is conditional: the framework can stand, but the predictive applications should be tempered or accompanied by independent validation and sensitivity analysis. The proposed concrete test would settle whether the Beta kernel is truly load-bearing or whether the predictions are robust to alternative kernels with the same first two moments.","tokens_in":14013,"tokens_out":6898,"duration_ms":75097,"concrete_test":"Run the CAR-T simulation of Section 4.3 with the same parameter values but replace the Beta kernel (14) by a truncated-normal or logistic-normal kernel with the same conditional mean and variance (same phi and eta); compare predicted relapse rates and CD34/CD19 distributions. If the outcomes change materially, the Beta kernel is load-bearing and the applications require empirical estimation of p. If the outcomes are insensitive, the unmeasured kernel does not threaten the central claim.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that Eq. (11) connects single-cell epigenetic state to population-level cancer dynamics, and that its hybrid implementations reproduce inflammation-induced tumorigenesis and CAR-T relapse, depends on a quantitatively correct inheritance kernel p(x,y). Section 3.2 explicitly concedes: \"biologically we cannot measure this function directly, and do not know the possible form neither.\" The paper then proposes a conditional Beta distribution, Eq. (14), whose shape parameters a(y)=eta(y)phi(y) and b(y)=eta(y)(1-phi(y)) are set by pre-defined functions phi(y) and eta(y). These functions are inherited from the author's own histone-modification simulation model [18,19], not from independent lineage-tracing or single-cell data. Since p(x,y) multiplies the entire renewal flux in Eq. (11), every prediction about heterogeneity, plasticity, tumorigenesis timing (Section 4.2), and CAR-T relapse (Section 4.3) is sensitive to the kernel. The paper presents no sensitivity analysis, no model-data comparison in this manuscript, and no code or data; the application sections refer to prior and preprint papers for details. The mathematical framework is coherent for arbitrary p, but the claimed biological reproductions are not yet supported by independent evidence.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript proposes a general mathematical framework for the evolutionary dynamics of heterogeneous and plastic stem cell populations, with applications to cancer. The authors first present a homogeneous G0 cell-cycle model, derive a delay differential equation, and then extend it to a delay integro-differential equation (Eq. 11) for the density of resting-phase stem cells with epigenetic state, where inheritance is described by a transition kernel. The kernel is specified as a conditional Beta distribution motivated by the authors' own histone-modification simulations. The framework is extended to genetic heterogeneity (Eq. 17) and to a local-state-transition approximation (Eq. 40). Two hybrid computational applications are described: inflammation-induced tumorigenesis and relapse after CAR-T therapy, with details referred to prior and preprint papers. The paper closes with four open mathematical problems: existence and uniqueness of steady states, entropy production and dissipation, inverse problems, and local transition approximations.","tokens_in":14373,"tokens_out":9463,"duration_ms":87977,"significance":"If the mathematical framework is accepted, it provides a coherent bridge between single-cell epigenetic state and population-level cancer dynamics, and the open problems may stimulate useful analysis. The derivation of Eq. (11) is a standard characteristic-line calculation and is clearly presented. However, the biological relevance of the framework hinges on the transition kernel, whose Beta form is not independently validated; the application results are not demonstrated in this manuscript. The framework is therefore more a proposal of a modeling paradigm than an established predictive model.","major_comments":[{"comment":"The general conditional Beta form is stated with exponents a(y)-1 and b(y)-1, but Eq. (14) omits the -1 terms. With a=eta*phi and b=eta*(1-phi), the density in Eq. (14) corresponds to Beta(eta*phi+1, eta*(1-phi)+1), whose mean is (eta*phi+1)/(eta+2) and whose variance is not phi*(1-phi)/(1+eta) as claimed. This inconsistency affects the parameterization used in the applications and must be corrected. Please specify whether the intended shape parameters are eta*phi and eta*(1-phi) (with exponents minus one) or eta*phi+1 and eta*(1-phi)+1.","section":"Section 3.2, Eq. (14)"},{"comment":"The transition kernel p(x,y) multiplies the entire renewal flux in Eq. (11), so all model predictions depend on its form. The paper states that 'biologically we cannot measure this function directly,' and the chosen Beta form is justified only by the authors' own simulations in references [18,19]. The application sections (inflammation-induced tumorigenesis and CAR-T relapse) do not present sensitivity analyses, quantitative model-data comparisons, or independent validation within this manuscript; they refer to prior and preprint papers for details. As a result, the claim that the framework reproduces these biological processes is not supported by the evidence presented here. The authors should either provide independent experimental support for the kernel, perform a systematic sensitivity analysis, or explicitly label the applications as conditional illustrations.","section":"Sections 3.2 and 4.2-4.3"},{"comment":"In the generalization to multiple epigenetic states, the text says 'we can extend the above Beta-distribution by the multiply rule p(x,y) = sum_i p_i(x_i,y).' A sum of marginal densities is not a joint density on the product space and does not satisfy the normalization condition unless the p_i are weighted categories. The intended independent-component joint distribution should be the product of the marginal densities. This point should be corrected to avoid ambiguity in applying the framework to high-dimensional states.","section":"Section 3.2, multi-dimensional extension"}],"minor_comments":[{"comment":"In the steady-state formula, the prefactor should be theta rather than eta; solving beta(Q*)=eta with beta from Eq. (5) yields Q* = theta [ (beta0+beta1-eta)/(eta-beta1) ]^{1/n}.","section":"Section 2, Eq. (7)"},{"comment":"The definition of total cell number is missing the differential: it should read Q(t) = integral over Omega of Q(t,x) dx.","section":"Section 5.2, Eq. (25)"},{"comment":"There are numerous typographical errors (e.g., 'challenge issue' and 'computations models' in the abstract, 'propose model' in Sections 4.1 and 6, 'serous' in Section 4.2, 'my promoted' in Section 4.3), and reference [46] is listed only as 'Preprint, 2019'; these should be corrected.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The manuscript relies heavily on the author's own prior work for the central equation (Eq. 11 from [22]), the Beta-kernel specification (from [18,19]), and both applications (from [13] and the preprint [46]). While the derivations in Sections 2-3 are sound, the paper is closer to a research-topic framework proposal than a self-contained predictive study. The editor may wish to consider whether the journal's scope is better suited to the open-problems framing or to a more substantiated applications paper."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Worth knowing: this is not a new-results paper. It is a framework-and-open-problems paper, and it mostly says so. The central equation (11) is the author's bioRxiv [22] and is closely related to the PNAS model [23]; the inflammation and CAR-T applications are condensed summaries of [13] and [46]. What is genuinely new here is the local-transition limit in Section 5.4 and the four mathematical problem statements. That is modest novelty, but the paper does not oversell it as a breakthrough.\n\nWhat it does well: the derivation of (11) from the age-structured model is a standard but clean characteristic-line calculation, and the notation is serviceable. More importantly, the paper is candid about its own weakest point: Section 3.2 explicitly says the transition function p(x,y) cannot be measured directly and that the possible form is unknown. That honesty matters. The open problems are also sensible and could be a useful starting point for someone entering this area.\n\nThe soft spots are real but proportionate. The load-bearing assumption is the conditional Beta kernel, and it comes from the author's own histone-modification simulations [18,19], not from independent lineage-tracing or single-cell data. Since p multiplies the whole renewal flux in (11), every prediction about heterogeneity, tumorigenesis timing, and CAR-T relapse is conditional on that kernel. The paper contains no sensitivity analysis and no model-data comparison in this manuscript. That is a serious limitation, though not a fatal one if the applications are read as illustrative rather than validated.\n\nThere are also algebraic slips. Equation (7) puts eta outside the bracket where theta should be; the correct steady state should have theta times the bracket, as the condition eta = beta0 theta^n/(theta^n + Q*^n) + beta1 gives directly. Equation (14), and the CAR-T kernel in Section 4.3, drop the -1 exponents from the general Beta form used in Section 3.2. Minor, but they need fixing. The local-transition derivation in Section 5.4 also deserves a careful check in review; the expansion of the convolution appears to mix the F term and the diffusion coefficient conventions.\n\nThe self-citations are not padding; they are the actual sources of the main equation and applications. But the paper should clearly mark which parts are prior results and which are new.\n\nBottom line: this is a paper for readers who want a single entry point to Lei's modeling framework and its open mathematical questions. It deserves a serious referee, yes, but as a framework/exposition paper with requested revisions, not as a primary research contribution. I would send it to review, ask the author to fix the algebra, mark prior results, and either provide code and data or temper the predictive claims.","headline":"A mostly synthetic framework paper that is honest about its open problems but leans heavily on the author's prior work, with the unmeasured Beta transition kernel as the main load-bearing weakness.","tokens_in":14881,"tokens_out":3296,"would_cite":true,"duration_ms":38552,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["92D25","45K05","92C50"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proposes that heterogeneous, plastic stem cell regeneration—captured by a delay integro-differential equation with an epigenetic inheritance kernel—is a general mathematical framework for cancer evolution.","keywords":["stem cell regeneration","differential-integral equation","epigenetic plasticity","cell heterogeneity","cancer development","computational cancer biology","CAR-T therapy relapse","open problems"],"falsifier":"Track individual labeled stem cells through one division in a relevant tissue (e.g., hematopoietic stem cells) by single-cell sequencing or reporter imaging, and compare the empirical daughter-state distribution for many mother states $y$ with the conditional Beta kernel (14) fitted through $\\varphi(y)$ and $\\eta(y)$. If the observed distributions are not Beta-shaped or the fitted shapes vary with mother state outside the assumed functional forms, the central kernel of Eq. (11) fails. A complementary experimental check: the CAR-T model predicts that combined CD19/CD123 CAR-T administration at specific ratios prevents CD19-positive relapse, a prediction testable directly in the mouse model that motivated the paper.","tokens_in":13786,"feed_emoji":"🧬","tokens_out":9046,"duration_ms":80626,"temperature":0.7,"pith_summary":"This paper proposes a general mathematical framework in which cancer evolution is driven by stem cells whose epigenetic state changes randomly at each division. Starting from a classical G0-phase cell-cycle model, the author derives a delay differential equation for a homogeneous stem cell pool, then generalizes it to a delay integro-differential equation, Eq. (11), in which cells are labeled by an epigenetic state $x$ and a kernel $p(x,y)$ gives the probability that a daughter cell of state $x$ comes from a mother of state $y$. The framework connects single-cell epigenetic state to population-level rates of proliferation, apoptosis, and differentiation, and it is extended to include genetic mutations in Eq. (17). Hybrid computational implementations of the framework are applied to inflammation-induced tumorigenesis and to relapse after CD19 CAR-T therapy, reproducing observed behavior in both settings. The author also identifies open mathematical problems—steady-state existence, entropy decomposition, inverse estimation of $p(x,y)$, and local state transitions—that must be solved before the framework can be used predictively.","feed_headline":"One equation links epigenetic shifts to cancer evolution","feed_subtitle":"A delay-integral equation with random epigenetic inheritance reproduces inflammation-driven tumors and CAR-T relapse.","key_machinery":"The load-bearing object is Eq. (11), a delay integro-differential equation for the resting-phase stem cell density $Q(t,x)$ over an epigenetic state space $\\Omega$. Its nonlocal term $$2\\int_\\$\\Omega$ \\$\\beta$(\\hat{Q}_{\\tau(y)},y)Q(t-\\tau(y),y)$e^{{-\\mu(y)\\tau(y)}}$p(x,y)\\,dy$$ carries the effect of cell division: a mother at state $y$ divides after duration $\\tau(y)$, suffers apoptosis at rate $\\mu(y)$, and produces daughters whose epigenetic state is drawn from the inheritance kernel $p(x,y)$. The kernel, named the transition function, is the mechanism that converts single-cell epigenetic plasticity into population heterogeneity. In the concrete one-dimensional model (14) it is a conditional Beta distribution with mother-state-dependent shape parameters $a(y)=\\eta(y)\\varphi(y)$ and $b(y)=\\eta(y)(1-\\varphi(y))$, fixed by the mean and variance functions $\\varphi(y)$ and $\\eta(y)$ obtained from the author's simulations of histone-modification inheritance. The characteristic-line integration that converts the age-structured PDE (10) into the delay equation (11) is the same technique that produces the homogeneous delay equation (4) from the G0 model; later, a local-transition approximation replaces the global kernel by a drift-and-diffusion expansion and yields the second-order equation (40).","core_discovery":"The central claim is that heterogeneous, plastic stem cell populations—the engine of cancer development—are governed by the delay integro-differential equation (11), in which the cell population $Q(t,x)$ at epigenetic state $x$ changes through loss by proliferation and differentiation and through influx from mothers of all states $y$ whose daughters randomly transition to $x$. The kernel $p(x,y)$ encodes epigenetic plasticity: it is the inheritance probability of daughter state given mother state. The paper argues that this equation places intrinsically nonlocal, non-measurable epigenetic transitions at the center of cancer dynamics and unifies single-cell gene-expression states, cell-cycle duration, apoptosis, differentiation, and feedback-dependent proliferation into one population-level description. In the one-dimensional stemness case (13) the kernel takes the explicit Beta-distribution form (14), and with genetic heterogeneity the framework becomes (17), a system over $m$ genetic types coupled by mutation probabilities. The same framework, discretized in hybrid single-cell simulations, reproduces the multi-pathway progression from chronic inflammation to malignancy and predicts that CAR-T-induced plasticity can drive CD19-positive relapse.","pith_inferences":["If the framework is right, the first practical task is not model refinement but measurement: single-cell lineage-tracing time series should be used to estimate $p(x,y)$ directly, and the inverse-problem formulation in Section 5.3 already sketches how that could be done.","The local-transition expansion (40) turns the global-kernel equation into a Fokker-Planck-type equation with drift $2a$ and diffusion $D$, suggesting a concrete bridge to classical mutation-selection balance and niche-construction models in evolutionary biology that the paper does not pursue.","The entropy problem has a testable corollary: if cancer development is an entropy-increasing process, then the entropy production rate under the framework should be large exactly during transitions from precancerous to malignant states, a quantity that could be estimated from sequentially sampled single-cell distributions.","The model's pathway-mutation treatment suggests a natural therapeutic extension not tested here: using the same hybrid equations to rank combinations of targeted agents by how far they push the population away from the malignant steady state, rather than by immediate cell kill."],"forward_implications":["In the homogeneous G0 model, abnormal growth requires at least one of three dysregulations—loss of differentiation/senescence, sustained proliferative signaling, or evasion of apoptosis—which the paper identifies with known cancer hallmarks.","With heterogeneity included, cell-to-cell variance in behavior is not noise layered on a homogeneous population but a consequence of random epigenetic inheritance, so therapy resistance and relapse become intrinsic outcomes of the same regenerative dynamics.","Gene mutations can be layered onto the framework through Eq. (17): each genetic type has its own kinetic rates and the mutation network $p_{i,j}$ couples the types, so mutation-driven intratumoral heterogeneity is a corollary of the same equation.","The applications imply that chronic inflammation can drive tumorigenesis through several distinct pathway combinations, and that CAR-T-induced plasticity toward hematopoietic stem-like and myeloid-like states is sufficient to produce CD19-positive relapse.","The four open mathematical problems—steady-state existence and stability, a nonnegative entropy production/dissipation decomposition, the inverse problem for $p(x,y)$, and local-transition limits—define what must be proved before the framework can be used for personalized prediction."],"supporting_citations":[{"why":"Supplies the original G0-phase age-structure model, restated as Eqs. (1)-(3), from which the delay equation (4) is derived.","marker":"[2]"},{"why":"Provides the characteristic-line reduction of the age-structured model to the delay differential equation (4).","marker":"[24]"},{"why":"Gives the general mathematical treatment of Eq. (11) and the explicit choices (15)-(16) used in applications.","marker":"[22]"},{"why":"Computational simulations of histone-modification inheritance that motivate the conditional Beta-distribution form of the transition kernel $p(x,y)$.","marker":"[18, 19]"},{"why":"Earlier discrete-state model with epigenetic transitions; supplies the steady-state eigenvalue analysis that the open problems extend.","marker":"[38]"},{"why":"Multiscale simulation application of the framework that reproduces inflammation-induced tumorigenesis and identifies the eight pathway mutations.","marker":"[13]"},{"why":"Single-cell-based model application to CD19 CAR-T therapy that reproduces relapse and generates the combined-CAR-T prediction.","marker":"[46]"},{"why":"Hallmarks of cancer used to interpret the G0 model's three routes to abnormal growth (apoptosis evasion, sustained signaling, differentiation loss).","marker":"[14]"}],"fun_headline_variants":["A delay-integral equation unifies cancer heterogeneity and plasticity","Random epigenetic transitions drive cancer population dynamics","One math model ties epigenetic states to tumor relapse","Equations for cancer's stem-cell plasticity and evolution"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole framework stands on the unmeasured inheritance kernel $p(x,y)$: the paper admits that biology cannot measure this function directly, and its concrete Beta-distribution form comes from the author's own simulations, so if real daughter-state distributions differ from that family every heterogeneous-model prediction built on Eq. (11) would shift.","fun_headline_variants_meta":{"raw":{"variants":["A delay-integral equation unifies cancer heterogeneity and plasticity","Random epigenetic transitions drive cancer population dynamics","One math model ties epigenetic states to tumor relapse","Equations for cancer's stem-cell plasticity and evolution"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000892,"raw_usage":{"total_tokens":3868,"prompt_tokens":991,"completion_tokens":2877,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":607,"completion_tokens_details":{"reasoning_tokens":2817}},"tokens_in":607,"tokens_out":2877,"duration_ms":21217,"temperature":1.0,"reasoning_tokens":2817,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:28:29.087325+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Track individual labeled stem cells through one division in a relevant tissue (e.g., hematopoietic stem cells) by single-cell sequencing or reporter imaging, and compare the empirical daughter-state distribution for many mother states $y$ with the conditional Beta kernel (14) fitted through $\\varphi(y)$ and $\\eta(y)$. If the observed distributions are not Beta-shaped or the fitted shapes vary with mother state outside the assumed functional forms, the central kernel of Eq. (11) fails. A complementary experimental check: the CAR-T model predicts that combined CD19/CD123 CAR-T administration at specific ratios prevents CD19-positive relapse, a prediction testable directly in the mouse model that motivated the paper.","supporting_citations":[{"cited_title":"On the existence of a G0-phase in the cell cycle","cited_arxiv_id":null,"evidence_quote":"Supplies the original G0-phase age-structure model, restated as Eqs. (1)-(3), from which the delay equation (4) is derived."},{"cited_title":"Multistability in an age-structured model of hematopoiesis: Cyclical neutropenia","cited_arxiv_id":null,"evidence_quote":"Provides the characteristic-line reduction of the age-structured model to the delay differential equation (4)."},{"cited_title":"A general mathematical framework for understanding the behavior of heterogeneous stem cell regeneration","cited_arxiv_id":null,"evidence_quote":"Gives the general mathematical treatment of Eq. (11) and the explicit choices (15)-(16) used in applications."},{"cited_title":"A mathematical model of stem cell regeneration with epigenetic state transitions","cited_arxiv_id":null,"evidence_quote":"Earlier discrete-state model with epigenetic transitions; supplies the steady-state eigenvalue analysis that the open problems extend."},{"cited_title":"Multiscale model- ing of inflammation-induced tumorigenesis reveals competing oncogenic and onco-protective roles for inflammation","cited_arxiv_id":null,"evidence_quote":"Multiscale simulation application of the framework that reproduces inflammation-induced tumorigenesis and identifies the eight pathway mutations."},{"cited_title":"Leukemic cell plasticity induces immune escape after CD19 chimeric antigen receptor T cell therapy of acute B lymphoblastic leukemia","cited_arxiv_id":null,"evidence_quote":"Single-cell-based model application to CD19 CAR-T therapy that reproduces relapse and generates the combined-CAR-T prediction."},{"cited_title":"The hallmarks of cancer","cited_arxiv_id":null,"evidence_quote":"Hallmarks of cancer used to interpret the G0 model's three routes to abnormal growth (apoptosis evasion, sustained signaling, differentiation loss)."}],"review_version":1}