{"id":"2d6f0cc1-7def-49b0-8032-cf3237c7ac38","arxiv_id":"2411.08512","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Noisy cell size control and volume partitioning modify the effect of growth-rate variability on population growth, and can make that variability beneficial when slow growers divide at smaller sizes.","lead":"A theoretical model shows how randomness in cell growth rates, birth sizes and division sizes determines the growth rate of a cell population. It finds that growth-rate variability can help population growth when slow-growing cells also divide at smaller sizes.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Eq. (30) is internally consistent, but its empirical applicability rests on excluding mother-daughter growth-rate correlations, which the paper itself notes can alter the conclusion.","rationale":"I traced the derivation of Eq. (30) for the adder special case with ā = 1, Sa = 0, and no size-control or partitioning noise, and reproduced the coefficient -1 + ln2/2 + Sb/(2 ln2) by expanding the Euler-Lotka equation and the tree birth-size moment equation to order CVα^2, including the mean birth-size correction. This confirms the mathematics is coherent. The potentially load-bearing restriction is the factorized kernel: all quoted predictions, including the beneficial threshold, are for uncorrelated growth rates. Because the paper explicitly acknowledges this and the reader's CONDITIONAL verdict already reflects the associated uncertainty, I recommend no change to the verdict. The proposed correlation-extension test would settle whether the qualitative claim survives in more realistic settings.","tokens_in":25332,"tokens_out":23431,"duration_ms":186860,"concrete_test":"Re-derive the cα coefficient in Eq. (30) using a correlated kernel ν(α|α') = ν(α)ν(α') [1 + ρ (α-⟨α⟩)(α'-⟨α⟩)/Var(α)], truncated to linear order in ρ, while keeping the same alpha-dependent linear map for a and b. Compute whether the sign of cα and the threshold āSa + (2-ā)Sb remain unchanged for ρ > 0. If the beneficial region disappears or shifts beyond the fitted E. coli values, the conclusion should be rephrased as valid only for uncorrelated growth rates; if it persists, the current claim stands.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central result Eq. (30) is derived from the factorized inheritance kernel Eq. (8), where the newborn growth rate is drawn from ν(α) independently of the mother's growth rate and of birth size. This factorization makes the tree marginal ρtree(α) = ν(α), so the coefficient cα = -1 + ln2/2 + (āSa + (2-ā)Sb)/(2 ln2) is a statement about uncorrelated single-cell growth rates. Real mother-machine data, including E. coli, show mother-daughter correlations in growth rates; the paper itself cites ref. [14] showing that such correlations can by themselves make fluctuations beneficial. If the true kernel is ν(α|α') rather than ν(α), the mean birth-size feedback in Eq. (29) and the Euler-Lotka weight both change, so the numerical threshold in Eq. (30) is not directly transferable to data without re-derivation. This is a stated limitation of the model rather than an internal inconsistency; the small-noise derivation itself is coherent and matches the simulations shown.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper develops a unified framework, based on a generalized Euler-Lotka equation (Eq. 6), to compute the steady-state population growth rate λ and the mean birth size in the tree statistics for cell populations with stochastic single-cell growth rates, noisy size control (noisy linear map, Eq. 11), and noisy volume partitioning (Eq. 9). The authors derive small-noise expansions for exponentially growing cells (Eqs. 16-17) and linearly growing cells (Eqs. 21-22), and they extend the model in Section III.C to allow the slope and intercept of the size-control map to depend on the growth rate (Eqs. 23-24). The central result is Eq. (30): λ/⟨α⟩ ≈ 1 − (1 − ln2/2 − (āSa + (2−ā)Sb)/(2 ln2)) CVα², so that growth-rate fluctuations can increase λ when slow-growing cells tend to divide at smaller sizes. The theoretical predictions are compared with agent-based simulations and show good agreement. An analysis of E. coli mother-machine data indicates that the size-control parameters do depend on growth rate, motivating the α-dependent model.","tokens_in":25557,"tokens_out":20090,"duration_ms":194987,"significance":"The paper provides a systematic and careful treatment of how different sources of single-cell noise affect population growth, extending earlier work by Lin and Amir and by Thomas. The result that size-control noise always increases λ for exponential growers, and that partitioning noise can be beneficial or detrimental depending on the control strategy, is new. The α-dependent size-control mechanism offers a distinct route for beneficial variability, separate from mother-daughter growth-rate correlations. The simulations are extensive and the expansions are tested beyond leading order; the recovery of known results as special cases (e.g., the asymmetric-division result of Barber et al.) lends confidence. The framework is general enough to cover linear and bi-linear growth laws, which broadens its applicability to organisms such as M. tuberculosis and S. pombe. If the identified limitations are addressed, the paper should be a useful contribution to quantitative cell biology.","major_comments":[{"comment":"The α-dependent parameterization a = ā(1 + Sa x) and b = b̄(1 + Sb x) is introduced without restricting the range of x. In the original linear map (11), the constraints b ≥ 0 and 0 ≤ a ≤ 2 are assumed. For the parameter values used to illustrate the beneficial effect in Fig. 5 (e.g., Sb = 3 with CVα up to 0.3), the added size b becomes negative for slow-growing cells (those with x < −1/Sb). A negative b implies sd < sb for the adder, so the generation time τd = ln(sd/sb)/α is not well-defined and the Euler-Lotka equation (6) no longer describes a meaningful cell cycle. The simulation results in Fig. 5 may therefore lie outside the model's domain of validity. The authors should either restrict the sensitivities so that b(α) > 0 and a(α) ∈ [0, 2) for all α in the support of ν (for example, use Sb values below the positivity bound 1/(2CVα) and choose Sb = 1.5 rather than Sb = 3 in the figure), or explicitly discuss the interpretation and consequences of negative b. This is directly relevant to the central claim because the beneficial regime is realized only for large positive S, where constraint violations are most likely.","section":"Section III.C, Eqs. (23)-(24), Fig. 5"},{"comment":"The factorized inheritance kernel K = ψ(sd|sb,α)ν(α)κ(sb|s'd) is the central structural assumption. It makes the tree distribution of growth rates equal to ν(α) and underlies the derivation of Eq. (30). The paper correctly notes in the Introduction that mother-daughter correlations in α are neglected and that such correlations can by themselves make fluctuations beneficial (ref. [14]). However, because real mother-machine data, including the E. coli data analyzed in Section C, exhibit such correlations, the quantitative threshold S > 2 ln2 − (ln2)² for beneficial fluctuations is not directly transferable to those data. I recommend that the authors (i) state this limitation explicitly in the abstract and conclusion, and (ii) ideally provide a short analysis or simulation for a correlated kernel ν(α|α′) to indicate how the coefficient cα changes. At minimum, the discussion should make clear that Eq. (30) applies to populations with uncorrelated single-cell growth rates.","section":"Section II.A, Eq. (8), and Section III.C, Eq. (30)"}],"minor_comments":[{"comment":"The phrase 'fluctuations in single-cell growth rates can be beneficial' in the abstract should be qualified with 'within our model, which neglects mother-daughter growth-rate correlations' to avoid overstatement, since the discussion already notes this limitation.","section":"Abstract and Discussion"},{"comment":"There is a typo: 'mircro-organisms' should be 'micro-organisms'.","section":"Section III.A"},{"comment":"The sentence 'on can also consider the situation' should read 'one can also consider the situation'.","section":"Section III.A"},{"comment":"The phrase 'allows use to identify' should be 'allows us to identify'.","section":"Section III.B"},{"comment":"In the text introducing the derivation, 'sensitivies' is a typo for 'sensitivities'.","section":"Supplementary Material, Section III"},{"comment":"The sentence 'This analysis also shows that the bounds on the population growth rate, ⟨α⁻¹⟩⁻¹ ≤ λ ≤ ⟨α⟩, derived in [10] when ψ(sd|sb,α)=ψ(sd|sb) are no longer valid in this case' could be clarified by adding 'in general' and explicitly noting that the lower bound ln2/⟨τd⟩fw still holds, which the paper does later.","section":"Section III.C"},{"comment":"For panels (a) and (b), the authors should state the order of the truncation used for the theoretical curves (first order in CVα²) and indicate which parameter combinations keep b(α) > 0 for all α in the plotted range.","section":"Figure 5"}],"recommendation":"major_revision","confidential_remarks":"The paper is technically sound and the derivations are careful, but the two major comments concern the domain of validity of the headline result. The positivity issue with b(α) in Section III.C is the more serious one: it affects the interpretation of Fig. 5 and the biological feasibility of the beneficial regime. The factorization limitation is acknowledged, but it should be made more prominent because it is load-bearing for the central claim. The E. coli data are used only to motivate the model, not to demonstrate the beneficial effect, which is appropriate; however, the title and abstract could be more cautious."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"What you should know: this is not another re-derivation of the growth-rate-variability penalty. The new content is the analytic coupling terms CV_alpha^2 CV_phi^2 and CV_alpha^2 CV_p^2 for exponential growers (eq 17), first-order size-control and partitioning corrections for linear growth (eq 22), and the alpha-dependent size-control route where growth-rate fluctuations can raise lambda (eq 30). The derivation from the generalized Euler-Lotka equation is careful, and the agent-based simulations in figs 2, 3 and 5 match the expansions, including the higher-order terms they keep for comparison. That is real evidence the expansions are right.\n\nCredit where due: the treatment of the forward-lineage birth-size distribution (eqs 13-14) is a clean trick, because it decouples size homeostasis from growth-rate noise and makes the tree statistics tractable. The recovery of the asymmetric-division result in the supplement is a good consistency check. The E. coli data analysis in section C is also a genuine step beyond prior work: the slope and intercept of the noisy linear map do appear to depend on growth rate, with slow growers closer to adder-like control.\n\nSoft spots, in proportion. First, the factorized kernel eq (8) is load-bearing: newborn alpha is drawn independently of the mother's alpha and of birth size. The author states this explicitly and notes in the introduction that mother-daughter correlations can already make fluctuations beneficial (ref [14]). So eq (30) is not directly transferable to mother-machine data where alpha correlations are present; it needs re-derivation with nu(alpha|alpha'). That is a stated scope limit, not an internal inconsistency, but it is the first thing a referee should probe. Second, everything is a small-noise expansion; the simulations use CV around 0.15 and match, but the paper does not map where the expansion breaks down, especially near the sizer for linear growth where the authors themselves note deviations in fig 3. Minor. Third, no simulation code or processed data are shipped. The simulation recipe is detailed enough to reproduce, but for a paper whose quantitative claims rest on expansions plus numerics, shipping code would settle it. Also minor.\n\nWho is this for: people working on microbial population fitness, size control, and Euler-Lotka theory. If you work on E. coli growth-rate noise and population growth, eqs (17) and (30) are directly usable.\n\nMy take: the central argument holds up. The beneficial-variability claim is conditional on the uncorrelated kernel, but conditional claims are allowed if stated, and it is stated. I would send this to a serious referee. Ask for code/data and for a paragraph on how nu(alpha|alpha') would shift eq (30).","headline":"A careful small-noise Euler-Lotka extension that finally couples growth-rate noise to size-control and partitioning noise; the math holds up, but the headline beneficial-variability regime rests on an uncorrelated-growth-rate kernel that the author explicitly flags.","tokens_in":26064,"tokens_out":2112,"would_cite":true,"duration_ms":21135,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["92D25","92C37"],"pacs":[],"model":"deepseek-v4-flash","headline":"Noise in single-cell growth rates can increase a population's growth rate when slow-growing cells divide at smaller sizes.","keywords":["population growth rate","Malthus parameter","cell size control","single-cell variability","Euler-Lotka equation","adder-sizer mechanisms","stochastic cell growth","coefficient of variation"],"falsifier":"Measure, in a mother-machine experiment, the joint distribution of mother and daughter single-cell growth rates: if the child growth rate is not independent of the mother's growth rate and birth size, the factorized kernel fails and eq. (30) is not the correct leading-order prediction. A second direct check is to measure $S_a$ and $S_b$ from linear fits of division size versus birth size in growth-rate bins, then compare population growth rates across conditions with different $\\text{CV}_\\alpha$; if $\\lambda$ does not rise once $\\bar{a}S_a + (2-\\bar{a})S_b$ exceeds $2\\ln 2 - (\\ln 2)^2$, the claimed beneficial effect is absent.","tokens_in":25104,"feed_emoji":"🦠","tokens_out":8932,"duration_ms":83180,"temperature":0.7,"pith_summary":"Single-cell experiments show that genetically identical cells differ in growth rate and in the sizes at which they are born and divide, and this paper asks how those fluctuations set the population growth rate—the Malthus parameter that serves as a fitness proxy. The author builds a stochastic population model whose inheritance law factorizes into a size-control kernel, a growth-rate distribution, and a volume-partitioning kernel, and derives closed-form small-noise expansions for the population growth rate and the population-level mean birth size. The headline result is that growth-rate variability can be beneficial: when the target division size depends on single-cell growth rate in the direction of slow-growing cells dividing at smaller sizes, the coefficient multiplying $\\text{CV}_\\alpha^2$ in $\\lambda/\\langle\\alpha\\rangle$ flips sign, so populations with more variable growth rates grow faster. The same formalism also shows that, for exponentially growing cells, noisier size control can partly offset the cost of growth-rate fluctuations and that the sizer mechanism outgrows the adder at fixed noise, while for linearly growing cells division-noise corrections are positive at first order. If correct, these formulas connect single-lineage measurements in mother machines to population-level fitness.","feed_headline":"Growth-rate noise can speed up a cell population","feed_subtitle":"The effect flips sign when slow growers divide small, says a model linking cell noise to population fitness.","key_machinery":"The engine of the paper is the generalized Euler–Lotka equation (eq. 6), the population analogue of the renewal equation that relates the distribution of birth, division, and growth-rate states in the population tree to the population growth rate $\\lambda$. The inheritance law entering it is assumed to factorize as $K(s_d,s_b,\\alpha|s'_d,s'_b,\\alpha') = \\psi(s_d|s_b,\\alpha)\\,\\nu(\\alpha)\\,\\kappa(s_b|s'_d)$, with $\\psi$ generated by the noisy linear map $s_d = a s_b + b + \\eta$ (eq. 11). Factorization implies that, in the forward-lineage statistics, birth size and growth rate are independent and that the forward birth-size distribution obeys a self-consistent equation; the author expands both sides of the Euler–Lotka equation in powers of the coefficients of variation and matches coefficients to get explicit formulas such as eqs. (17), (22), and (30). The parameter $a$ indexes the size-control strategy—0 is a sizer, 1 is an adder, 2 would be a timer—and appears explicitly in every coefficient, which is how the paper compares mechanisms.","core_discovery":"On its own terms, the paper establishes that the population growth rate $\\lambda$ is, to leading order in the coefficients of variation, a sum of squares of single-cell noise: growth-rate noise $\\text{CV}_\\alpha^2$, size-control noise $\\text{CV}_\\phi^2$, volume-partitioning noise $\\text{CV}_p^2$, and, in the $\\alpha$-dependent case, the sensitivities $S_a$ and $S_b$ of the slope and intercept of the noisy linear map $s_d = a s_b + b + \\eta$. For exponentially growing cells the central formula is $\\lambda/\\langle\\alpha\\rangle \\approx 1 - \\bigl(1 - \\tfrac{\\ln 2}{2} - \\tfrac{\\bar{a}S_a + (2-\\bar{a})S_b}{2\\ln 2}\\bigr)\\,\\text{CV}_\\alpha^2$ (eq. 30). When $\\bar{a}S_a + (2-\\bar{a})S_b$ is large enough, the correction is positive and fluctuations in single-cell growth rates increase population growth; the author argues this is the regime where slow-growing cells divide at smaller sizes than fast-growing cells. For the previously studied case where size control does not depend on growth rate, the paper recovers the known detrimental effect and adds new cross-terms: noise in size control is always beneficial, while noise in volume partitioning is beneficial only below $a \\approx 0.97$, so the sizer is faster than the adder for fixed noise levels. The same Euler–Lotka machinery, applied to linear growth, yields first-order positive corrections from size-control and partitioning noise even when growth rates are identical across cells.","pith_inferences":["A natural next test is to measure the joint distribution of mother and daughter growth rates in the same experiments; the factorized-kernel assumption (eq. 8) is the main place where real cells could depart from the model, and the paper itself notes that mother–daughter correlations can already make fluctuations beneficial.","If eq. (30) survives data, it reframes the evolution of cell size control: variability in growth rate would not be a nuisance to be suppressed but a trait that can be selected for, provided slow growers divide small—an argument the author does not develop into an evolutionary claim.","The linear-growth result suggests that for organisms with linear single-cell growth, measurement of size-control noise $\\text{CV}_\\phi$ alone is enough to predict a fitness advantage of noisy division, without requiring growth-rate noise; this is a sharper, testable prediction than the exponential case.","The same expansion method could be applied to growth laws with more than two phases or to time-varying growth rates; the supplementary treatment of bi-linear fission yeast growth indicates the machinery extends, but the paper leaves those extensions for future work."],"forward_implications":["For exponentially growing cells with $\\alpha$-independent size control, the population growth penalty from growth-rate noise is reduced by size-control noise; at fixed $\\text{CV}_\\phi$, $\\text{CV}_p$, and $\\text{CV}_\\alpha$, the sizer yields a larger $\\lambda$ than the adder or timer-like mechanisms.","Noise in division-size control is never harmful at leading order for exponential growth, while noise in volume partitioning helps only for $a \\lesssim 0.97$ and is essentially neutral for the adder.","For linearly growing cells, noise in size control and partitioning raise $\\lambda$ even when all cells share the same growth rate, with the adder being the slowest strategy.","When $a(\\alpha)$ and $b(\\alpha)$ are extracted from single-cell data, the model predicts that growth-rate variability can push $\\lambda$ above the value for a uniform population, and the bound $\\lambda \\le \\langle\\alpha\\rangle$ can fail while $\\lambda \\ge \\ln 2/\\langle\\tau_d\\rangle_{\\mathrm{fw}}$ survives.","Mother-machine estimates of $\\text{CV}_\\alpha$, $\\text{CV}_\\phi$, and $\\text{CV}_p$ are directly usable in these expansions to predict population-level fitness from lineage data."],"supporting_citations":[{"why":"Supplies the baseline result that growth-rate noise alone gives $\\lambda/\\langle\\alpha\\rangle \\approx 1 - (1-\\ln 2/2)\\text{CV}_\\alpha^2$ and the bounds $\\langle\\alpha^{-1}\\rangle^{-1} \\le \\lambda \\le \\langle\\alpha\\rangle$.","marker":"[10]"},{"why":"Provides the prior derivation of how growth-rate fluctuations interact with cell size control and the numerical observation that added-size noise in the adder increases population growth.","marker":"[12]"},{"why":"Shows that mother-daughter correlations in single-cell growth rates can already make fluctuations beneficial, which motivates the author's factorization assumption and its limits.","marker":"[14]"},{"why":"Supplies the mother-machine E. coli data from which the author binned cells by growth rate and fitted the slope and intercept of the noisy linear map to obtain $S_a$ and $S_b$.","marker":"[7]"},{"why":"Gives the forward-lineage birth-size coefficient-of-variation formula that the paper extends and uses to close the expansion for the tree statistics.","marker":"[33]"},{"why":"Introduces the noisy linear map $s_d = a s_b + b + \\eta$ that defines the sizer, adder, and intermediate size-control mechanisms used throughout the model.","marker":"[8]"},{"why":"Provides the prior formula for population growth with asymmetric division under a perfect sizer, which the supplementary material recovers as a special case of this framework.","marker":"[15]"}],"fun_headline_variants":["Slow growers dividing small make growth noise beneficial","Cell noise can flip from harmful to helpful for populations","Growth-rate noise helps populations when slow cells divide small","Model shows when cell noise accelerates population growth","Noise in cell growth rate can benefit population growth"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole calculation rests on the assumption that a newborn cell's growth rate is drawn independently of its mother's growth rate and of the size it was born with, encoded in the factorized inheritance kernel of eq. (8); if mother–daughter growth-rate correlations are present, the predicted size and sign of the growth-rate effect would need to change.","fun_headline_variants_meta":{"raw":{"variants":["Slow growers dividing small make growth noise beneficial","Cell noise can flip from harmful to helpful for populations","Growth-rate noise helps populations when slow cells divide small","Model shows when cell noise accelerates population growth","Noise in cell growth rate can benefit population growth"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000258,"raw_usage":{"total_tokens":1651,"prompt_tokens":1081,"completion_tokens":570,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":697,"completion_tokens_details":{"reasoning_tokens":498}},"tokens_in":697,"tokens_out":570,"duration_ms":6044,"temperature":1.0,"reasoning_tokens":498,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T21:32:33.261843+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure, in a mother-machine experiment, the joint distribution of mother and daughter single-cell growth rates: if the child growth rate is not independent of the mother's growth rate and birth size, the factorized kernel fails and eq. (30) is not the correct leading-order prediction. A second direct check is to measure $S_a$ and $S_b$ from linear fits of division size versus birth size in growth-rate bins, then compare population growth rates across conditions with different $\\text{CV}_\\alpha$; if $\\lambda$ does not rise once $\\bar{a}S_a + (2-\\bar{a})S_b$ exceeds $2\\ln 2 - (\\ln 2)^2$, the claimed beneficial effect is absent.","supporting_citations":[{"cited_title":"2021 Modeling the impact of single-cell stochasticity and size control on the population growth rate in asymmetrically dividing cells","cited_arxiv_id":null,"evidence_quote":"Provides the prior formula for population growth with asymmetric division under a perfect sizer, which the supplementary material recovers as a special case of this framework."}],"review_version":1}