{"id":"3fd489ff-107a-4323-a1f4-95709a597f5a","arxiv_id":"1908.05050","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"A Keller-Segel model extended with chemokinesis predicts that speed responses to attractants can enhance bacterial accumulation, split populations, and reduce navigation errors in decaying gradients.","lead":"This paper builds a mathematical model of bacteria that both swim toward food and speed up when food is more concentrated, then uses it to predict how such bacteria gather around nutrient sources. The model predicts faster, stronger accumulation than chemotaxis alone, plus new patterns like split populations and broader migration waves.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Factor-2 error in chemokinetic drift magnitude (Eq. 3 vs Eq. 2 and 10c) invalidates threshold (15) and the n=40 split in Fig. 4; corrected threshold is ~79.6, not 39.8.","rationale":"The reader's conditional verdict was based on the unmeasured Hill response function and instantaneous adaptation. Those concerns are real but acknowledged by the authors, and the supplementary scans show the predicted trends are not tied to a single parameter value. My concern is different and more immediate: the microscopic derivation in the paper itself fixes the coefficient of V_k, and that coefficient is inconsistent with the diffusion coefficient in Eq. 2 and with Eq. 10c. The two-state derivation gives V_k = -D (∂_x v)/v = -v/(2α)∂_x v, not -v/α ∂_x v. This is not a matter of external calibration; it is a factor-2 error inside the model. It changes the central threshold (15), and therefore undermines the specific demonstration of the qualitative 'split population' effect in Fig. 4. It also changes all quantitative predictions, including the transient-source bias mitigation magnitude. The paper otherwise has strengths: a transparent derivation, clearly stated assumptions, shared code and data, and analytical steady states. A correction of the factor and a rerun of the affected figures is a bounded, concrete fix; if the corrected results preserve the qualitative effects for step-like responses, the paper can remain a useful conditional contribution.","tokens_in":18095,"tokens_out":13264,"duration_ms":141909,"concrete_test":"Independently re-derive Eq. 3 by integrating the velocity-jump equations for R and L with α_R=α_L=α to check the coefficient, then rerun the shared Zenodo code with the corrected V_k (factor 1/2 in Eq. 11d) and a consistent D0_b = v0^2/(2α) for the Fig. 4 parameter set. Confirm whether the n=40 curve still develops the split and recompute the threshold; a threshold near ~79.6 and disappearance of the n=40 split would settle the concern.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"In Sec. II, the model defines D_b = v^2/(2α) (Eq. 2) but V_k = -(v/α) ∂_x v (Eq. 3). For the two-state run-and-tumble process with α_R = α_L = α, the diffusion limit gives J = -(v^2/(2α))∂_x b - (v/(2α))(∂_x v)b, so the chemokinetic drift should be half as large. The inconsistency is internal: Eq. (10c) writes the diffusive term as -(v^2/α)∂_x b, again a factor of 2 larger than Eq. 2, and the non-dimensionalisation in Sec. IIC uses D0_b = v0^2/α instead of v0^2/(2α). Because V_k enters Eq. (11d) directly, condition (15) is off by the same factor: it should be n > 8 δ0 ω Kχ/(ω+Kχ)^2 (1/η + 1/2), not 4... . For the parameters of Fig. 4 (δ0=50, η=2, ω=0.2, Kχ=0.53), the threshold becomes ≈79.6 instead of 39.8, so the n=40 curve labelled 'just above the threshold' is actually below it. The two-subpopulation split in the fixed gradient is advertised as the principal qualitative modification; with the corrected factor it is not demonstrated for n=40, and the biological Hill estimates n≈1–3 cited from [5,13] are even further from the threshold. The effect could still exist for n→∞ step-like responses, but the quantitative boundary and the illustrative figure must be corrected.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript extends the Keller-Segel model to describe bacterial populations that combine chemotaxis with chemokinesis, i.e., a swimming speed that depends on local attractant concentration. The model is derived from run-and-tumble dynamics and yields a drift-diffusion flux with a concentration-dependent diffusion coefficient, a chemokinetic drift directed down speed gradients, and a chemotactic drift whose sensitivity grows with the square of the swimming speed. The authors apply the model to three scenarios: a fixed linear attractant gradient, a self-generated gradient on an agar plate, and a transient diffusing source. They report that chemokinesis can split the population in a fixed gradient, broaden a traveling wave, and reduce the navigation bias caused by a decaying attractant source. The paper includes analytical threshold conditions and numerical solutions, with code and data deposited on Zenodo.","tokens_in":18521,"tokens_out":17517,"duration_ms":159797,"significance":"The paper addresses a relatively underexplored aspect of bacterial motility and provides a transparent continuum framework that connects microscopic parameters (swimming speed, tumble rate) to macroscopic chemotactic sensitivity. The explicit inclusion of temporal-gradient effects following Hein et al. and the treatment of speed-dependent diffusion and chemotactic sensitivity are useful contributions. The quantitative predictions are falsifiable with existing microfluidic and agar-plate assays, and the authors are honest about the lack of measured chemokinetic response functions. The availability of code and data strengthens reproducibility. However, a factor-2 error in the chemokinetic drift and diffusion terms changes the key threshold condition and undermines the illustrative demonstration of the population split, so the results cannot be accepted in their current form.","major_comments":[{"comment":"The qualitative predictions depend on the assumed Hill-type chemokinetic response function v(c) = v0 + vc c^n/(c^n + kc^n) (Eq. 9), whose parameters η, ω, and n are not measured but are estimated only loosely from literature values (Appendix B gives n ≈ 1–3 from refs. [5,13]). After correcting the flux factor error, the threshold for the population split becomes n ≳ 80 for the parameters used in Fig. 4, which is far above the cited empirical estimates. This raises the question of whether the predicted split is a realistic biological phenomenon or an artifact of an extreme parameter choice that was not flagged as such in the abstract. The authors should provide a more systematic robustness analysis (e.g., non-monotonic or much shallower response functions, and different relative values of ω and Kχ) or explicitly present the split as a high-steepness limiting case. The Discussion acknowledges the missing measurements, but the central claims would be considerably strengthened by making the sensitivity to the assumed response function more explicit and by testing the predictions against alternative functional forms.","section":"Sec. IIB, Eq. (9); Appendix B; Sec. IV"}],"minor_comments":[{"comment":"The angle brackets used in the definition of I2(t) appear to denote path averaging, but the notation is not defined here and the placement of the brackets is ambiguous; please clarify by adding an explicit definition and consistent notation.","section":"Appendix A, Eq. (A14)"},{"comment":"The main text states that parameters were chosen 'to best illustrate the chemokinetic effect' and Appendix B notes that η=2 is 'an extreme value'. Please state this explicitly in the main text where Fig. 3 and Fig. 4 are discussed, so that readers do not interpret the chosen parameters as representative of real bacteria.","section":"Sec. IIC and Appendix B"},{"comment":"The phrase 'measuring bias' in the abstract is unclear; consider 'navigation bias' or 'measurement bias' to convey that the temporal change in the attractant gradient reduces the accuracy of chemotactic navigation.","section":"Abstract and Sec. IIIC"},{"comment":"The text refers to Fig. S2 for the comparison of chemotactic populations at constant speed, but the main-text relevant figure is Fig. 5; please ensure the cross-references are consistent and that the reader can follow which results are in the main text versus the supplementary material.","section":"Sec. IIIB, Fig. 5 and S2"}],"recommendation":"major_revision","confidential_remarks":"The factor-2 error in the flux is a serious internal inconsistency that must be fixed: it changes the threshold condition and the illustrative figure supporting the paper's central qualitative claim. Once corrected, the authors should also reconsider how strongly they state the qualitative predictions, given the large gap between the corrected threshold and the literature-based estimates of the Hill coefficient. The paper is within scope for the journal and, if revised carefully, could make a useful contribution."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The stress-test note is right: the factor-2 discrepancy is real and it affects the paper's main quantitative threshold. But the paper deserves credit first. It is the first continuum Keller-Segel treatment of chemokinesis combined with chemotaxis, and it adds a sensible temporal-gradient correction to the chemotactic drift, derived from run-and-tumble dynamics in the de Gennes style. The authors are transparent about what is known and what is assumed: they explicitly state the chemokinetic response function v(c) has not been measured, and they use a Hill form as a placeholder. Code and data are on Zenodo, and parameters are documented. The steady-state solution in Eq. (17) is a nice, compact result.\n\nNow the problem. In Eq. (2) the diffusion coefficient is D_b = v^2/(2α), but the flux in Eq. (10c) uses v^2/α for the diffusive term, and the chemokinetic drift in Eq. (3) is V_k = -(v/α)∂_x v. The standard two-state derivation (Schnitzer, Othmer–Hillen) gives a chemokinetic drift of -(v/(2α))∂_x v, a factor of two smaller. The non-dimensionalisation also uses D0_b = v0^2/α, which is inconsistent with the paper's own Eq. (2). This shifts the threshold in (15) by the same factor: for the parameters in Fig. 4, the correct threshold is about 79.6, not 39.8, so the n=40 curve labelled \"just above the threshold\" is actually below it. That part of the text and figure needs re-running. It also means the two-subpopulation split, advertised as a key qualitative result, is not demonstrated for parameters in the biologically plausible Hill range (n=1–3). The effect may still appear for very steep, step-like responses (n→∞), but the quantitative boundary is wrong.\n\nThe other qualitative predictions—wave broadening in self-generated gradients and mitigation of the transient-source bias—do not depend on this threshold and should survive the fix. The paper's central claim that chemokinesis can modify chemotaxis qualitatively is probably correct. The response function being assumed is not a fatal flaw; the authors are upfront about it and even argue for measurements. Instantaneous speed adaptation is also acknowledged.\n\nOverall, a solid modeling paper with a genuine quantitative error in the central flux. It deserves peer review: the idea is good, the derivation in Appendix A is mostly careful, and the error is fixable. I'd ask the authors to correct the factor 2, update the threshold and Fig. 4, and re-state which predictions are robust. I would not cite the current version because of the error, but I would cite the corrected one.","headline":"A useful continuum model for chemotaxis-chemokinesis with a factor-2 error in the drift that shifts a headline threshold but leaves the qualitative conclusions intact.","tokens_in":19017,"tokens_out":6142,"would_cite":false,"duration_ms":60402,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Chemokinesis combined with chemotaxis can split a bacterial population in a fixed gradient, broaden its traveling wave, and reduce the navigation bias from a decaying attractant source.","keywords":["bacterial chemotaxis","chemokinesis","run-and-tumble motility","Keller-Segel model","population dynamics","traveling waves","transient nutrient gradients","microfluidic assays"],"falsifier":"Measure $v(c)$ for a chemokinetic bacterium while holding the gradient fixed, then run the model with the measured function: if a two-front split appears in a fixed linear gradient for a Hill coefficient below the threshold in Eq. (15), or if no split appears for a coefficient above it, the central drift-competition claim is falsified. A separate check is to observe whether the traveling-wave broadening disappears when the speed response is shallow, since the model predicts broadening only when the speed gradient is significant over the band.","tokens_in":17919,"feed_emoji":"🦠","tokens_out":9518,"duration_ms":89682,"temperature":0.7,"pith_summary":"This paper extends the standard continuum model of run-and-tumble bacterial migration to include chemokinesis, the speed-up of swimming in higher attractant concentrations. It argues that the combination of chemotaxis and chemokinesis changes population behaviour qualitatively, not just quantitatively: in a fixed gradient part of the population can drift toward low attractant, in a self-generated gradient the travelling band broadens, and around a decaying nutrient pulse the migration is faster and less biased by the pulse's decay. These predictions matter because many soil and marine bacteria display chemokinesis, and the extra drift could explain why some species lack the sharp bands seen in Escherichia coli. If correct, chemokinesis is not a small correction to chemotaxis but a competing migration mechanism that can dominate under steep speed responses.","feed_headline":"Bacteria that speed up near food can split a migrating swarm in two","feed_subtitle":"The same speed response also broadens traveling waves and steadies navigation around decaying nutrient pulses.","key_machinery":"The central object is the bacterial flux $$J = -D \\partial b/\\partial x + V_k b + V_\\chi b,$$ where the diffusion coefficient $D = v^2/(2\\alpha)$ depends on the local swimming speed $v$, the chemokinetic drift $V_k = -(v/\\alpha)\\partial v/\\partial x$ points down the speed gradient, and the chemotactic drift $V_\\chi = \\chi(\\nabla f_\\chi + (1/v)\\partial_t f_\\chi)$ points up the attractant gradient. Here $\\chi(x) = \\chi_0 v(x)^2/v_0^2$, so faster swimming raises chemotactic sensitivity, and the speed obeys $v(x)=v_0 + v_c c^n/(c^n+k_c^n)$. The competition between $V_k$ and $V_\\chi$ carries the argument: when the Hill coefficient $n$ exceeds a threshold set by Eq. (15), chemokinetic drift dominates at $c=k_c$ and the population can accumulate at low attractant instead of high.","core_discovery":"The paper's central claim is that for a population combining chemotaxis and positive chemokinesis, the total bacterial flux decomposes into contributions from diffusion, a chemokinetic drift, and a chemotactic drift, and the chemokinetic term can qualitatively reshape the accumulation. In a fixed linear gradient, the model's criterion for chemokinetic drift to dominate is $n > 4\\delta_0\\omega K_\\chi/(\\omega+K_\\chi)^2(1/\\eta+1/2)$ evaluated at $C=\\omega$; when met, part of the population is carried toward low attractant instead of continuing up-gradient. In a self-generated gradient, the travelling wave of a chemotactic-chemokinetic population is faster but broader than the pure chemotactic wave, and the pulse travels slower than a uniformly fast-swimming population because front speed scales with pulse amplitude. Around a transient diffusing source, chemokinesis leads to faster, stronger accumulation and reduces the bias caused by the attractant's decay, because the temporal term in the chemotactic drift is weighted by $1/v$.","pith_inferences":["If the measured speed-vs-attractant curve for a species is steeper than the threshold, Eq. (15) becomes a testable diagnostic: species with steep responses should show the split, species with shallow responses should not, regardless of absolute speed increase.","The model's assumption of instantaneous speed adaptation sets a lower bound on the effect; real adaptation delays of order 10–200 s would likely smear the split and further broaden the wave, so transient measurements could discriminate.","The same flux decomposition applies to synthetic chemokinetic swimmers whose speed rises with fuel concentration, so the predicted low-fuel accumulation should be reproducible with Janus particles in a fuel gradient.","The metabolic-cost discussion implies that chemokinesis concentrates swimming expenditure just where nutrients are available, which is a plausible evolutionary reason for the behavior even when it slows the band's leading edge."],"forward_implications":["In fixed microfluidic gradients, a steep enough chemokinetic response (large $n$) should be observable as a temporary or persistent split, with a subpopulation trapped near the half-saturation concentration rather than at the source.","In agar plate assays, chemokinetic species should show broad, less sharply peaked rings rather than the sharp Adler bands, and their population should grow faster in the same nutrient budget.","Around a decaying nutrient pulse, chemotaxis-chemokinesis should accumulate a larger fraction of cells at the source than pure chemotaxis, and the prediction should be insensitive to whether the temporal term in the drift is included.","When the chemokinetic response is very steep (step-like), chemokinesis can inhibit rather than help chemotaxis, so the relation between speed response steepness and accumulation is non-monotonic."],"supporting_citations":[{"why":"provides the continuum random-walk derivation that yields the chemokinetic drift $V_k=-(v/\\alpha)\\partial v/\\partial x$","marker":"[25]"},{"why":"supplies the microscopic run-and-tumble calculation relating chemotactic sensitivity to $v^2/\\alpha$ and the memory kernel","marker":"[30]"},{"why":"motivates the temporal term $\\partial_t f_\\chi/v$ in the chemotactic drift for decaying gradients","marker":"[28]"},{"why":"provides the agar-plate self-generated gradient model, logistic growth, and the standard parameter set used in simulations","marker":"[29]"},{"why":"reports the step-like chemokinetic response in a coral pathogen and forms the baseline agent-based model the continuum results are compared with","marker":"[5]"},{"why":"gives speed-dependent chemotactic precision data and adaptation times for marine bacteria used for parameter motivation","marker":"[15]"},{"why":"describes light-controlled swimming speed in E. coli, proposed as the experimental route to test the model's predictions","marker":"[33]"},{"why":"supplies the result that traveling pulse speed scales with pulse amplitude, used to explain the slower chemokinetic pulse","marker":"[35]"}],"fun_headline_variants":["Speeding up near food can drift bacteria away from the source","Chemokinesis splits swarms and steadies navigation around decaying food","Bacteria that swim faster in food can reverse their migration direction","Chemokinesis broadens waves and reduces bias from decaying attractants"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The weakest load-bearing premise is the assumed functional form of the chemokinetic response, $v(c)=v_0+v_c c^n/(c^n+k_c^n)$ with instantaneous adaptation: the pure chemokinetic speed increase as a function of attractant concentration has never been systematically measured for any species, and if the true response is non-monotonic, much shallower, or has a very different half-saturation, the predicted split, band broadening, and bias reduction would weaken or disappear.","fun_headline_variants_meta":{"raw":{"variants":["Speeding up near food can drift bacteria away from the source","Chemokinesis splits swarms and steadies navigation around decaying food","Bacteria that swim faster in food can reverse their migration direction","Chemokinesis broadens waves and reduces bias from decaying attractants"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000197,"raw_usage":{"total_tokens":1384,"prompt_tokens":986,"completion_tokens":398,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":602,"completion_tokens_details":{"reasoning_tokens":325}},"tokens_in":602,"tokens_out":398,"duration_ms":4600,"temperature":1.0,"reasoning_tokens":325,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:26:48.944841+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure $v(c)$ for a chemokinetic bacterium while holding the gradient fixed, then run the model with the measured function: if a two-front split appears in a fixed linear gradient for a Hill coefficient below the threshold in Eq. (15), or if no split appears for a coefficient above it, the central drift-competition claim is falsified. A separate check is to observe whether the traveling-wave broadening disappears when the speed response is shallow, since the model predicts broadening only when the speed gradient is significant over the band.","supporting_citations":[{"cited_title":"Functional analysis of nine putative chemoreceptor proteins in Sinorhizobium meliloti","cited_arxiv_id":null,"evidence_quote":"provides the continuum random-walk derivation that yields the chemokinetic drift $V_k=-(v/\\alpha)\\partial v/\\partial x$"},{"cited_title":"The chemokinetic and chemo- tactic behavior of Rhodobacter sphaeroides: Two inde- pendent responses","cited_arxiv_id":null,"evidence_quote":"supplies the microscopic run-and-tumble calculation relating chemotactic sensitivity to $v^2/\\alpha$ and the memory kernel"},{"cited_title":"5, n = 1, T = 0","cited_arxiv_id":null,"evidence_quote":"motivates the temporal term $\\partial_t f_\\chi/v$ in the chemotactic drift for decaying gradients"},{"cited_title":"Chemoki- nesis in Rhodobacter sphaeroides is the result of a long term increase in the rate of flagellar rotation","cited_arxiv_id":null,"evidence_quote":"provides the agar-plate self-generated gradient model, logistic growth, and the standard parameter set used in simulations"},{"cited_title":"A bacterial pathogen uses dimethyl- sulfoniopropionate as a cue to target heat-stressed corals","cited_arxiv_id":null,"evidence_quote":"reports the step-like chemokinetic response in a coral pathogen and forms the baseline agent-based model the continuum results are compared with"},{"cited_title":"Conversely, if η < 0 (i.e","cited_arxiv_id":null,"evidence_quote":"gives speed-dependent chemotactic precision data and adaptation times for marine bacteria used for parameter motivation"},{"cited_title":"Speed-dependent chemotactic precision in marine bacteria","cited_arxiv_id":null,"evidence_quote":"describes light-controlled swimming speed in E. coli, proposed as the experimental route to test the model's predictions"},{"cited_title":"Chemosen- sory behavior in protozoa","cited_arxiv_id":null,"evidence_quote":"supplies the result that traveling pulse speed scales with pulse amplitude, used to explain the slower chemokinetic pulse"}],"review_version":1}