{"id":"a95947c2-6162-4bd1-9069-a979a75cb864","arxiv_id":"1908.10564","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":11,"one_line_summary":"The authors show that a generalized parameter-cascades method can extract hepatic transport rates and compartment-specific fluorescence scaling factors from noisy intravital microscopy data.","lead":"This paper adapts a statistical fitting method to estimate how fast a fluorescent dye moves through liver compartments in living rats from microscopy videos. It then applies the method to show that chronic kidney disease slows dye uptake into liver cells and a bile-blocking drug stops dye entry into bile channels.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Quantitative rates depend on prespecified α, α′, kHT; the insensitivity claim is asserted without a sensitivity analysis, so the central quantitative claim remains conditional on untested external inputs.","rationale":"The central claim is that parameter cascades, generalized to include compartment-specific measurement parameters, yields quantitative transport rates from IVM fluorescence. This requires the inferred rates to be determined by the data rather than by arbitrary fixed inputs. The paper fixes α, α′, and kHT to resolve structural unidentifiability and then asserts insensitivity in a single sentence without demonstration. Because Eq. (15) shows the observed sinusoid and hepatocyte signals are linear mixtures of unmodified and glucuronidated species, there is a genuine scaling trade-off: a perturbed α/α′ changes how much signal is attributed to S versus S′, and kHT directly feeds H′. The synthetic validation in Tables 1–3 is real evidence that the optimizer recovers known parameters when the fixed measurement parameters are exactly known, but it does not test the mismatch between fixed values and the actual liver environment. Thus the quantitative case-study values are conditional in a way the text does not quantify. The proposed grid re-estimation is a direct computational check that would settle the concern. If it passes, the central claim is substantially supported; if it fails, the paper should be reframed as qualitative or re-fit with independently measured α, α′, and kHT. I therefore keep the reader's CONDITIONAL verdict.","tokens_in":19008,"tokens_out":9648,"duration_ms":104700,"concrete_test":"Re-run the SI Appendix B estimation procedure on the experimental traces underlying Tables 4–6 across a grid of prespecified values, e.g., α ∈ {0.25, 0.5, 0.75}, α′ ∈ {0.1, 0.25, 0.5}, and kHT ∈ {0.1, 0.5, 1.0} min−1, or over literature ranges from refs. 74–77. For every combination, record all 12 inferred rates. The concern is settled if (1) each rate stays within the reported standard deviations, and (2) the sham-versus-5/6N ordering of kS→H and k′S→H and the TLC kH→C near-zero result are preserved at all grid points. If any rate moves outside the reported uncertainty or an ordering flips, the insensitivity claim is falsified and the quantitative conclusions must be re-scoped.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's quantitative output is conditional on three externally fixed parameters: α=0.5, α′=0.25, and kHT=0.5 min−1. The text asserts that 'Our quantitative conclusions are insensitive to exact parameter estimates used initially for kHT, α and α′' but provides no sensitivity analysis, profile, or table in the main text. This assertion is load-bearing because the measurement model in Eq. (15) makes yS a weighted sum αS+α′S′, so a different α/α′ directly reallocates sinusoid signal between unmodified and glucuronidated fluorescein; similarly, kHT controls the source term for H′ and can trade against k′S→H and k′H→S. The identifiability analysis in SI Appendix B is cited to justify fixing these parameters, and refs. 74–77 are cited to show the chosen values are experimentally accessible, but neither source demonstrates that the inferred rates in Tables 4–6 are invariant to the chosen values. If the insensitivity assertion is false, the reported kS→H decrease in 5/6N rats and the near-zero kH→C in TLC-treated rats could reflect the fixed inputs rather than the biology.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a parameter-cascade method to infer kinetic and measurement parameters of ordinary differential equation models for hepatic fluorescein transport from intravital microscopy (IVM) fluorescence time traces. The model couples unmodified and glucuronidated fluorescein species across three compartments (sinusoid, hepatocyte, canaliculus), with measurement parameters mapping hidden species concentrations to observed fluorescence and accounting for compartment-specific quenching. The method is validated on synthetic data from a two-state model, a generalized two-state model with two species, and the FitzHugh-Nagumo model, with increasing noise levels. The authors then apply the method to previously published IVM data from sham, 5/6-nephrectomized (5/6N), and taurolithocholate (TLC)-treated rats, reporting reduced sinusoid-to-hepatocyte uptake in 5/6N and near-zero canalicular efflux in TLC. Identifiability requires prespecifying α=0.5, α′=0.25, and kHT=0.5 min−1, and the paper asserts without demonstrating that the results are insensitive to these choices.","tokens_in":19398,"tokens_out":3801,"duration_ms":38580,"significance":"If the method holds, it offers a route to quantitative in vivo transport rates from fluorescence data alone, avoiding absolute concentration calibration. The inclusion of measurement parameters in a parameter-cascade framework is a genuine methodological extension, and the synthetic validation on multiple nonlinear systems supports the inference machinery. The application to disease models addresses a physiologically important problem. However, the central quantitative claims for the real data rest on three externally fixed parameters, and the insensitivity assertion is not backed by a sensitivity analysis. Additionally, the printed objective functions omit the measurement matrix, making the method as stated internally inconsistent. With these points addressed, the paper would be a useful contribution to quantitative intravital microscopy and ODE parameter estimation.","major_comments":[{"comment":"The objective functions used in the inner and outer optimization loops are written as sums of squared differences between the data y_i and the spline approximation xhat_i of the state vector. This is inconsistent with the measurement model in Eq. (2), y = Hx + ε, because y_i are fluorescence measurements in three compartments while x_i are six hidden species concentrations. As written, minimizing |y_i − xhat_i|^2 would only be correct if H were the identity, which contradicts the entire purpose of estimating measurement parameters. The equations must be revised to include H, e.g., |y_i − H xhat_i|^2, and the dimensions of H clarified (Eq. (2) says H is m×n but Eq. (15) shows it is 3×6). Without this correction, the method as described is not reproducible and the statistical meaning of the outer criterion is unclear.","section":"Materials and Methods, Eqs. (4)-(5)"},{"comment":"The identifiability analysis (cited to SI Appendix B) forces the authors to fix α=0.5, α′=0.25, and kHT=0.5 min−1. The text states that 'our quantitative conclusions are insensitive to exact parameter estimates used initially for kHT, α and α′' but gives no sensitivity analysis, profile likelihood, or supporting table. This is load-bearing: the sinusoid measurement is yS = αS + α′S′, so changing α/α′ directly reallocates the observed sinusoid signal between unmodified and glucuronidated fluorescein; similarly, kHT enters the source term for H′ and can trade off against k′S→H and k′H→S. If the reported differences between sham, 5/6N, and TLC (e.g., kS→H 2.24 vs 2.07 min−1, kH→C 0.0032 min−1) are sensitive to these fixed inputs, the biological conclusions would be compromised. The authors should either provide a systematic sensitivity analysis over a physiologically plausible range of these parameters and show that the inferred rates and their confidence intervals are stable, or explicitly state the dependence and weaken the conclusions.","section":"Results, Tables 4-6 and the paragraph on prespecified parameters"}],"minor_comments":[{"comment":"In the differential equation for dH′/dt, the term representing loss to the canaliculus is written as k_H→C H′ rather than k′_H→C H′. This is inconsistent with the model schematic and with the dC′/dt equation, which uses k′_H→C. Please correct this apparent typo.","section":"Eq. (14)"},{"comment":"The paper contains several typographical errors, including 'Microcopy' in the title, 'reducig' in the Results text, 'mehods' in the Discussion, and 'These coe cients' in the duplicated passage near the end. A careful proofreading pass is needed.","section":"General typos"},{"comment":"In Algorithm 1, the inner optimization is written as minimizing Jin(c′|θ′, λ) using a prime on c that is not defined elsewhere; Eq. (4) also uses c without saying whether it is the vector of all spline coefficients. Please clarify the notation.","section":"Algorithm 1 and Eq. (4)"},{"comment":"The experimental tables report standard deviations for the estimated parameters, but the main text does not explain how these are computed. The SI is cited for confidence intervals, but a brief statement in the main text about the procedure (e.g., bootstrap or asymptotic covariance) would improve interpretability.","section":"Standard deviations in Tables 4-6"}],"recommendation":"major_revision","confidential_remarks":"The manuscript uses previously published IVM data from a co-author's group (Ref. 29) and prespecified parameters motivated partly by co-author Day's fluorescence lifetime work (Ref. 77). This is not inherently problematic, but the novelty as a methods paper would be stronger if the new experimental contribution (if any) and the exact role of the reused data were stated explicitly. The sensitivity-analysis requirement is the main gate for acceptance."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nQuick take: this is a credible applied inference paper, not a methods breakthrough. The new piece is modest but real: they include compartment-specific measurement parameters as structural parameters in a parameter-cascades fit, and they push the framework through a six-species, three-observation hepatic transport model with three experimental conditions. The synthetic validation (two-state, generalized two-state, FitzHugh-Nagumo) is the strongest part of the paper. It shows the machinery recovers known ground truth at several noise levels, with error bars that grow sensibly as noise increases. That is real evidence the core estimator works.\n\nThe experimental conclusions are biologically plausible and align with prior physiology: 5/6N nephrectomy lowers sinusoid-to-hepatocyte uptake modestly, and TLC nearly abolishes canalicular efflux. The comparison of sham vs. 5/6N rates (kS→H 2.24 vs. 2.07 min−1) is not huge relative to the reported standard deviations, so I would not oversell that difference. The TLC effect on kH→C (0.0032 min−1) is stark and clearly consistent with the known mechanism.\n\nSoft spots, in proportion:\n\n1. The printed objective functions in Eqs. (4)–(5) omit the measurement matrix H. As written they compare yi to x̂i, but y is length r (here 3) and x̂ is length m (here 6). That is dimensionally inconsistent unless the hat implicitly means Hx̂. This is a typo-level fix, but in a methods paper it matters, and it is the kind of thing a referee should catch.\n\n2. The insensitivity claim is load-bearing and unsupported. They prespecify α=0.5, α′=0.25, kHT=0.5 min−1 and say the conclusions are insensitive, but no sensitivity analysis appears in the main text or in what I can see of the SI. Because the sinusoid measurement is a weighted sum αS+α′S′, and kHT feeds the H′ source term, these choices can trade against the inferred rates. The stress-test note is right: this is the central quantitative vulnerability. It may well be that the rates are robust in practice, but the paper needs to show a profile or a small table demonstrating that before I trust Tables 4–6.\n\n3. Minor: confidence intervals and identifiability details are deferred to SI. That is fine for a JPC B paper, but it does mean the main text's precision claims are not self-contained.\n\nOverall, the method is honestly presented, the synthetic work is reproducible in spirit, and the biology is not overinterpreted. The missing sensitivity analysis is addressable and should be the main referee request. I would take this paper seriously and engage with it, but I would not cite the absolute rate values from the real-data tables until the sensitivity question is settled.\n\nRecommendation: send to peer review. The synthetic validation alone justifies that. Ask for the H fix, a sensitivity table for α, α′, kHT, and a more explicit statement about how the reported standard deviations were computed. After that, it is publishable.\n\nBest,\n[You]","headline":"A legitimate adaptation of parameter cascades to hepatic IVM transport, with solid synthetic validation but a load-bearing insensitivity claim that is asserted, not demonstrated.","tokens_in":19911,"tokens_out":780,"would_cite":false,"duration_ms":9886,"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":"Intravital microscopy fluorescence traces, modeled with compartment-specific measurement parameters and inferred by a generalized parameter-cascades scheme, can yield quantitative liver transport rates without invasive calibration.","keywords":["intravital microscopy","hepatic transport","parameter cascades","ODE parameter estimation","measurement parameters","fluorescein glucuronide","chronic kidney disease","taurolithocholate"],"falsifier":"Re-run the full hepatic-transport inference on the same sham, 5/6N, and TLC traces with $\\alpha$, $\\alpha'$, and $k_{HT}$ varied across their physiologically plausible ranges (for example $k_{HT}$ from 0.1 to 2 min$^{-1}$, and $\\alpha$ and $\\alpha'$ from 0.1 to 0.9). If the reported differences between sham and 5/6N uptake rates, or the near-zero canalicular rates in TLC, move by more than the quoted standard deviations, the central conclusion is not robust.","tokens_in":18809,"feed_emoji":"🔬","tokens_out":8093,"duration_ms":74907,"temperature":0.7,"pith_summary":"This paper tries to establish that quantitative transport kinetics across liver compartments can be recovered from intravital microscopy fluorescence data alone. Because fluorescein emits differently in the sinusoid, hepatocyte, and canaliculus, and because both fluorescein and its glucuronide form contribute to the signal, the authors add compartment- and species-specific measurement parameters to a six-species ODE transport model and estimate those parameters together with the kinetic rates. The estimation tool is a generalized parameter-cascades procedure that avoids numerically solving the ODEs. Validation on simulated two-state, two-species, and a benchmark nerve-axon excitation model shows the rates are recoverable, and on real rat-liver data the method finds that 5/6 nephrectomy lowers sinusoid-to-hepatocyte uptake while taurolithocholate nearly abolishes hepatocyte-to-canaliculus efflux. If these estimates are right, live-imaging studies could quantify drug effects on specific hepatic transport steps without invasive sampling.","feed_headline":"Fluorescence video alone yields liver transport rates","feed_subtitle":"Compartment-aware ODE fits show kidney disease slows hepatic uptake while taurolithocholate blocks bile-side efflux","key_machinery":"The load-bearing device is a generalized parameter-cascades estimator. In parameter cascades, the ODE solution is approximated by a B-spline expansion whose coefficients (nuisance parameters) are fit in an inner optimization under a roughness penalty that enforces fidelity to the ODE, while the structural parameters—the kinetic rates—are updated in an outer optimization. The paper's generalization puts the entries of the measurement matrix $H$ into the structural parameter vector, so the per-compartment fluorescence visibility fractions ($\\alpha$, $\\beta$, $\\gamma$ and their primed counterparts) are inferred from the same data instead of being calibrated separately. Identifiability analysis is used to decide which parameters must be prespecified: $\\alpha = 0.5$, $\\alpha' = 0.25$, and $k_{HT} = 0.5$ min$^{-1}$, with the remaining twelve parameters estimated.","core_discovery":"The central claim is that the hidden species concentrations in each liver compartment and the measurement parameters that map them to observed fluorescence are jointly identifiable from noisy time traces, provided three parameters are fixed from outside knowledge. Implemented for hepatic transport, the model has six species—fluorescein and glucuronidated fluorescein in sinusoid, hepatocyte, and canaliculus—coupled by linear rates, with three fluorescence measurements described by a rectangular measurement matrix. On sham-control rat data the inferred sinusoid-to-hepatocyte uptake rate is $k_{S\\to H} = 2.24$ min$^{-1}$; in 5/6 nephrectomy it drops to $2.07$ min$^{-1}$, while in taurolithocholate-treated rats the hepatocyte-to-canaliculus rate collapses to $k_{H\\to C} = 0.0032$ min$^{-1}$ and canalicular loss to $k_C = 0.0026$ min$^{-1}$, matching the known cholestatic action of TLC. The paper also demonstrates on synthetic data that the estimation scheme recovers ground-truth parameters even when measurement noise is added.","pith_inferences":["A natural extension, not carried out in the paper, is to replace the fixed $\\alpha$, $\\alpha'$, and $k_{HT}$ values with per-animal measurements from fluorescence lifetime imaging; this would turn the sensitivity question into a direct comparison rather than an assumption.","If the measurement matrix is identifiable from data, the scheme could generalize to spectrally distinct probes or to transporters with multiple substrates, using the same latent-species structure.","The model's linear kinetics could be replaced by saturable transport kinetics to capture transporter saturation; a testable prediction is that high-dose fluorescein should slow apparent uptake rates when uptake is saturable.","Comparing transporter-knockout animals with the same inference pipeline would test whether each transporter's removal shifts the corresponding edge rate constant, a prediction the current data does not resolve."],"forward_implications":["Transport-step-specific rate constants can be assigned to disease or drug effects from image time series, not just from isolated-cell or vesicle assays.","Because measurement parameters are estimated, the method should work on data with different optical conditions or quencher environments, and on other fluorescent probes that change emission by compartment.","The near-zero canalicular rates recovered under taurolithocholate give a quantitative, image-based signature of cholestasis that could be tracked over time or across doses.","For chronic kidney disease, the specific reduction in sinusoid-to-hepatocyte uptake offers a concrete kinetic explanation for impaired hepatic clearance previously reported in 5/6N rats."],"supporting_citations":[{"why":"Supplies the parameter-cascades optimization that the paper generalizes to estimate measurement parameters together with kinetic rates.","marker":"(38)"},{"why":"Defines the quantitative intravital microscopy approach and the sinusoid-hepatocyte-canaliculus compartments used for the model.","marker":"(21)"},{"why":"Provides the previously published sham, 5/6N, and TLC intravital data sets and the chronic kidney disease transport finding that the study re-analyzes.","marker":"(29)"},{"why":"Documents fluorescein glucuronidation after intravenous injection, motivating the six-species model with two chemical forms.","marker":"(30)"},{"why":"Supplies the structural identifiability analysis used to decide which parameters must be prespecified.","marker":"(69)"},{"why":"Establishes that taurolithocholate blocks hepatic organic anion transport, the pharmacological expectation the TLC case study tests.","marker":"(80)"}],"fun_headline_variants":["Fluorescence video quantifies liver transport","Liver transport rates decoded from video","Video-based model tracks liver drug kinetics","Kidney disease slows liver uptake in video model","TLC cholestasis captured by compartment-aware model"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the three prespecified values taken from earlier studies—the two sinusoid visibility fractions ($\\alpha=0.5$, $\\alpha'=0.25$) and the hepatocyte glucuronidation rate ($k_{HT}=0.5$ min$^{-1}$)—are accurate enough that the inferred rates are insensitive to their exact values, an assertion the paper states without reporting a sensitivity analysis.","fun_headline_variants_meta":{"raw":{"variants":["Fluorescence video quantifies liver transport","Liver transport rates decoded from video","Video-based model tracks liver drug kinetics","Kidney disease slows liver uptake in video model","TLC cholestasis captured by compartment-aware model"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001376,"raw_usage":{"total_tokens":5589,"prompt_tokens":975,"completion_tokens":4614,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":591,"completion_tokens_details":{"reasoning_tokens":4549}},"tokens_in":591,"tokens_out":4614,"duration_ms":37610,"temperature":1.0,"reasoning_tokens":4549,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T10:40:13.522467+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-run the full hepatic-transport inference on the same sham, 5/6N, and TLC traces with $\\alpha$, $\\alpha'$, and $k_{HT}$ varied across their physiologically plausible ranges (for example $k_{HT}$ from 0.1 to 2 min$^{-1}$, and $\\alpha$ and $\\alpha'$ from 0.1 to 0.9). If the reported differences between sham and 5/6N uptake rates, or the near-zero canalicular rates in TLC, move by more than the quoted standard deviations, the central conclusion is not robust.","supporting_citations":[],"review_version":1}