{"id":"b6f2ac12-2216-4c55-bd5a-930878495562","arxiv_id":"2509.06113","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A mother-machine style microfluidic device for suspension T-cells enables direct lineage tracking and reveals symmetric division and sizer-like growth control in Jurkat cells.","lead":"Researchers built a microfluidic chip that traps and tracks single non-adherent T-cells over several generations, something that is difficult with standard methods. The device lets them measure how cells divide, grow, and time their division, and shows cell division in this leukemia line is symmetric with a sizer-like growth control.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Biological headline likely reflects a curated subpopulation: manual channel selection with no exclusion counts could bias the division-time distribution and the m=-0.24 sizer slope.","rationale":"The reader's weakest-assumption is exactly the selection bias in the manually curated channel set, and it is the load-bearing point: every quantitative biological claim (division symmetry, sizer slope, division-time peak) is computed only from channels that already passed a 'successful division' filter. I did not find a separate flaw that would overturn the device contribution; the engineering claims are supported by detailed fabrication, geometry, and CFD. The concern does not move the verdict because the reader already assigned CONDITIONAL on essentially these grounds. I would add that the lack of sample sizes and confidence intervals is a second facet of the same problem, and that the abstract's 'slightly asymmetric' wording is in tension with the reported symmetric Gaussian, but neither replaces selection bias as the central issue.","tokens_in":14675,"tokens_out":6113,"duration_ms":59279,"concrete_test":"Require the authors to release the per-channel census (total channels imaged, number manually selected, and number excluded for each stated reason) and to recompute Fig. 4a-c using all channels that contain at least one cell and pass a pre-registered automated quality criterion, with per-cell data and bootstrap 95% intervals for mu, sigma, m, and the division-time peak. If the selected fraction is low, the sizer slope shifts by more than its bootstrap width, or the division-time peak moves away from ~22 h, the population-level growth-strategy claim is not established.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section IV (Image segmentation) states that the first analysis step is manual selection of channels 'where a single cell successfully enters and divides'; the explicit exclusions are multi-cell entries, non-dividing/senescent cells, deaths, overlapping daughters, and fused cells. Every population-level quantity in Fig. 4 — the symmetric-division Gaussian (mu=0.5, sigma=0.05), the regression slope m=-0.24, and the 22 h division-time peak — is computed only from this curated set. The excluded categories are precisely the non-proliferating or abnormal cells, so the reported numbers describe a selected, potentially fast-proliferating subpopulation rather than the population growth strategy claimed in the title and abstract. No counts of included/excluded channels, no per-reason exclusion tallies, and no sample sizes or confidence intervals are reported, so it is impossible to determine whether the selected channels are representative or whether m=-0.24 is statistically distinguishable from 0 (adder) or from -1 (pure sizer). In addition, the abstract's phrase 'slightly asymmetric volume division' contradicts the reported single symmetric Gaussian centered at 0.5; if the authors intend asymmetry, the presented fit does not support it. The device engineering is independently supported by fabrication detail and CFD, but the biological conclusions are underdetermined by the evidence shown.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript presents a mother-machine-like microfluidic device for non-adherent mammalian cells, specifically Jurkat T-cells. The device uses open-ended trapping channels, dual-inlet loading, and flow-through perfusion, with geometry optimized through computational fluid dynamics. The authors demonstrate lineage tracking over multiple generations and report three biological measurements: an inherited-fraction distribution fit by a single Gaussian with mu=0.5 and sigma=0.05 (interpreted as symmetric division), a regression slope of m=-0.24 between birth size and added size (interpreted as a sizer-like strategy), and a division-time distribution peaking at about 22 h. The paper is primarily an engineering and platform contribution, with the biological findings presented as direct single-cell measurements.","tokens_in":15035,"tokens_out":4359,"duration_ms":40540,"significance":"If the biological findings are robust, the platform is valuable because direct lineage tracking of suspension cells is technically challenging, and the device design is documented in sufficient detail to be reproduced. Strengths include the open-ended channel geometry, the dual-inlet loading scheme, the CFD-assisted optimization, and the direct time-lapse tracking rather than indirect flow-cytometry inference. However, the population-level biological claims currently rest on a manually curated subset of channels with no reported exclusion counts, sample sizes, or confidence intervals. The device engineering may be sound, but the growth-strategy conclusions are underdetermined by the evidence as presented.","major_comments":[{"comment":"The first analysis step is a manual selection of only channels where a single cell enters and divides, with explicit exclusions of multi-cell entries, non-dividing/senescent cells, death, overlapping daughters, and fused cells. Because every population-level quantity in Fig. 4—the Gaussian fit, the slope m=-0.24, and the division-time peak at 22 h—is computed only from this curated set, the reported numbers describe a potentially fast-proliferating subpopulation rather than the population growth strategy claimed in the title and abstract. No counts of included or excluded channels, no per-reason exclusion tallies, and no total number of tracked lineages are reported. Please provide the full channel census and a sensitivity analysis showing how the main statistics change under alternative inclusion rules.","section":"Section IV, Image segmentation; Fig. 4"},{"comment":"The paper reports a Gaussian fit with mu=0.5 and sigma=0.05 and a regression slope of m=-0.24 without any sample size, confidence interval, or goodness-of-fit statistic. The text states that a single timelapse permits following 'hundreds of division events', but the actual n for each panel is never stated. Without a confidence interval, the slope m=-0.24 cannot be distinguished from 0 (adder) or from -1 (pure sizer); the claim of a 'sizer-like' strategy is therefore not statistically grounded. Please provide n, standard errors, p-values, and fit diagnostics for every panel in Fig. 4, including the growth-rate distributions in Fig. 5.","section":"Fig. 4"},{"comment":"The abstract states that cells exhibit 'a slightly asymmetric volume division', but Section II.D reports that the fit returned a single Gaussian with mu=0.5 and sigma=0.05, explicitly 'consistent with symmetric divisions'. A single Gaussian centered at 0.5 is symmetric, so the abstract overstates the evidence. Please reconcile the abstract with the reported fit, or provide a separate measure of asymmetry (e.g., the two-Gaussian fit mentioned in the text) that supports the word 'asymmetric'.","section":"Abstract and Section II.D"},{"comment":"The regression slope m=-0.24 is presented as the key evidence for a sizer-like strategy, but the analysis uses area at mid-height rather than volume, and the text acknowledges that area and volume are nonlinearly related. The line is fit to normalized areas without reporting the regression method, the uncertainty on the slope, the scatter around the line, or the number of cells. Please report the correlation coefficient, the standard error of m, and the results of testing against the null hypothesis m=0, so that the reader can assess the strength of the size-control claim.","section":"Section II.D and Fig. 4b"}],"minor_comments":[{"comment":"There are several typographical errors: 'fatc' should be 'fact', 'proceeeded' should be 'proceeded', and 'Figure b4' should be 'Figure 1b4'.","section":"Throughout"},{"comment":"The caption says 'three of the six distinct lineages identified', but the total number of lineages in the full experiment is never defined. Please clarify what 'six' refers to and how the displayed branches were chosen.","section":"Fig. 3c caption"},{"comment":"Please state the exact number of division events, cell cycles, and channels used for each panel of Fig. 4, rather than the qualitative 'hundreds of division events'.","section":"Section II.D"},{"comment":"The claim that cells in 45-degree channels have the highest growth rates is made without sample sizes or a statistical test. Please add the number of cells per condition and a significance test for the differences among inclinations.","section":"Fig. 5b"},{"comment":"For a quantitative single-cell study, 'available from the corresponding author upon reasonable request' is not ideal. Consider depositing the segmentation masks, lineage tables, and analysis code in a public repository to enable reproduction and reuse.","section":"Data Availability"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is best framed as a device paper; the biological claims in the abstract currently exceed the evidence. The validation against Refs. [5, 47, 48] relies mostly on measurements from the same group, so the confirmatory value is limited. I would not reject on engineering grounds, but the population growth-strategy conclusions must either be properly quantified (with channel census, sample sizes, and confidence intervals) or substantially softened. The contradiction between the abstract's 'slightly asymmetric' and the Results' 'symmetric' fit should be fixed before resubmission."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThis paper is worth a look if you care about microfluidics for suspension cells. The device itself — open-ended mother machine channels, dual-inlet loading, three channel inclinations for Jurkat cells — is a genuine, useful engineering step. The fabrication detail is solid. CFD simulations back the trapping-efficiency claims. That part holds up.\n\nThe biological conclusions are softer. The paper reports symmetric division (single Gaussian at 0.5), a sizer-like regression slope m=-0.24, and a division-time peak at 22h. But these numbers rest on a manually curated subset of channels: only channels where a single cell enters and divides successfully are analyzed. The excluded categories — multiple cells, non-dividers, deaths, overlaps, fusion — are exactly the cells you would need to count for a population-level growth strategy. No exclusion counts, no sample sizes, no error bars, no confidence intervals for the slope. So m=-0.24 could be anywhere, and the claim that it indicates a sizer-like strategy is underdetermined. The abstract's phrase \"slightly asymmetric volume division\" also overstates the data, which the text itself says is a single symmetric Gaussian centered at 0.5. That mismatch should be fixed.\n\nThere is no data or code release; the Data Availability line says available from the corresponding author on reasonable request. For a methods paper, shipping the analysis pipeline would add a lot of confidence.\n\nThe citation pattern is fine. The biological results match the authors' own prior indirect measurements, which is confirmatory, not circular — they are measuring directly here. But it does mean the \"new biology\" is limited.\n\nWho is this for? Groups building single-cell tracking devices for non-adherent mammalian cells. They will get real value from the design and the CFD approach. As a biology paper, it needs much tighter statistics. As an engineering paper, it is solid.\n\nI would send it to peer review, with the expectation that the biological claims get constrained or the analysis gets substantially more rigorous. A desk reject would be too harsh.","headline":"A solid device paper whose biological claims need more rigorous statistics before they can carry the title.","tokens_in":15455,"tokens_out":1444,"would_cite":true,"duration_ms":12908,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"T-cells tracked in traps split evenly, then partly size-correct","keywords":["microfluidics","mother machine","suspension cells","T cells","lineage tracking","cell size control","sizer model","division symmetry"],"falsifier":"A reader could count every channel that contained a cell, record why each was excluded, and recompute the division-time distribution and the added-size versus birth-size regression with the excluded channels included. If the peak moves above 22 h or the slope moves toward zero, the reported sizer strategy is an artifact of channel selection.","tokens_in":1471,"feed_emoji":"🧫","tokens_out":2004,"duration_ms":77286,"temperature":0.7,"pith_summary":"This paper reports a microfluidic device that adapts the mother-machine design, narrow channels that trap one founder cell and let its progeny pile up in a line, to suspension cells such as human leukemia T-cells, which normally float and are hard to track. Continuous flow through open-ended channels keeps the cells supplied with nutrients, and imaging every minute for four days lets the authors reconstruct lineage trees across multiple generations. From those lineages they report three quantitative results: division is largely symmetric (a single Gaussian peak at an inherited fraction of 0.5 with $\\sigma=0.05$), cells follow a partially compensating sizer-like strategy (a regression slope of $m=-0.24$ between added size and birth size), and division times peak at about 22 hours. The device also shows that channel inclination and medium flow change proliferation, with 45-degree channels giving the highest growth rates and no-flow conditions roughly doubling division times. If the measurements hold, the device gives suspension-cell biology a direct single-cell window that previously required indirect population-level inference.","feed_headline":"T-cells tracked in traps split evenly, then partly size-correct","feed_subtitle":"Lineage tracking of floating T-cells shows symmetric division and partial size compensation, previously only inferred.","key_machinery":"The central object is the open-ended mother-machine microchannel, a cell-sized trap (about 15 µm wide and 20 µm tall in this version) whose distal end connects to a drainage channel through a narrow constriction. The paper's design differs from closed-end bacterial mother machines by opening the distal end, which creates a slow unidirectional flow through each trap. That flow does two jobs: it carries cells in while preventing escape, and it continuously exchanges nutrients so cells deep in the channel are not starved. The main flow channel also has staggered pillar arrays that break up aggregates, and a dual-inlet layout separates buffer flow from cell loading. Around this geometry, the analysis machinery is a deep-learning segmentation pipeline that turns time-lapse images into cell masks and lineage trees, from which birth sizes, division sizes, and division times are extracted.","core_discovery":"On its own terms, the paper's central claim is that an open-ended, flow-through mother-machine chip can confine non-adherent mammalian cells well enough to follow hundreds of divisions and extract lineage statistics, and that human leukemia T-cells tracked this way divide symmetrically and control size through a partial sizer mechanism. The evidence is direct: the inherited-volume-fraction distribution is a single Gaussian centered at 0.5 with standard deviation 0.05, the regression of added size on birth size has slope $m=-0.24$ (between timer-like $m>0$ and pure sizer $m=-1$), and interdivision times peak near 22 h, consistent with culture doubling times. The paper also reports that 45-degree channel inclination gives the highest trapping probability (about 60%) and the fastest growth, and that stopping medium flow shifts division times to a bimodal distribution with peaks near 28 h and 46 h. These are presented as direct measurements of proliferation features that earlier work had inferred from flow cytometry.","pith_inferences":["A reanalysis that reports how many channels were rejected, and why, would test whether the 22 h peak and $m=-0.24$ slope are representative of the proliferating population or biased toward channels with fast, clean divisions.","The paper measures cell cross-sectional area, not volume; if a volume-resolved version produced a more negative slope, the underlying control could be closer to a pure sizer than the area-based number suggests.","The no-flow result implies that cells born under good conditions retain a memory of those conditions; this could be tested directly by switching flow on and off at defined times and watching division times in the same lineage.","The device could be adapted to primary T-cells or to drug perturbation, where lineage-level information about asymmetric division or cell-cycle arrest would be clinically relevant; that extension is natural but not demonstrated here."],"forward_implications":["Direct lineage reconstruction becomes available for suspension cells, not just adherent cells or bacteria, so growth models can be tested on single cells rather than inferred from population snapshots.","The measured symmetric division and partial sizer slope provide a single-cell check on earlier cytometry-based estimates for this leukemia cell line.","Device geometry can be tuned: 45-degree channels combine higher trapping with faster growth, making them the preferred configuration for proliferation studies.","Continuous medium flow is a functional requirement, not a convenience; without it, division times widen and later-born cells divide much more slowly.","Because channels are sized to the cell, the same design should transfer to other suspension cell types by rescaling dimensions."],"supporting_citations":[{"why":"Supplies the original mother-machine architecture that this device adapts for suspension cells.","marker":"[36]"},{"why":"Gives the indirect flow-cytometry estimate of a size-dependent division strategy that the paper's slope $m=-0.24$ directly confirms.","marker":"[5]"},{"why":"Provides the deep-learning segmentation method used to obtain cell masks and lineage statistics.","marker":"[59]"},{"why":"Supports the dual-inlet configuration that prevents cell accumulation near the loading port.","marker":"[42]"},{"why":"Underpins the definitions of birth and division sizes used to compute added size and division symmetry.","marker":"[45]"},{"why":"Supports the timing convention for measuring size before cytokinesis.","marker":"[46]"},{"why":"Documents how PDMS mother-machine devices prolong division times, the baseline the paper compares its 22 h peak against.","marker":"[40]"},{"why":"Provides a prior cultivation platform for mammalian suspension cells with single-cell resolution, the comparison point for this device.","marker":"[58]"}],"fun_headline_variants":["T-cell trap tracks divisions, shows symmetric split and size fix","Non-adherent cells tracked over generations: symmetric division, partial sizer","Mother machine for floating cells reveals symmetric division and size control","Suspension cells in microchannels divide symmetrically, then partly size-correct"],"cache_read_input_tokens":17664,"weakest_assumption_plain":"The load-bearing premise is that the channels manually selected as containing one cleanly dividing cell are representative of the proliferating population; if slow-dividing or unhealthy cells are preferentially excluded, the measured division-time peak and sizer slope would be biased.","fun_headline_variants_meta":{"raw":{"variants":["T-cell trap tracks divisions, shows symmetric split and size fix","Non-adherent cells tracked over generations: symmetric division, partial sizer","Mother machine for floating cells reveals symmetric division and size control","Suspension cells in microchannels divide symmetrically, then partly size-correct"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000624,"raw_usage":{"total_tokens":2880,"prompt_tokens":929,"completion_tokens":1951,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":545,"completion_tokens_details":{"reasoning_tokens":1874}},"tokens_in":545,"tokens_out":1951,"duration_ms":15000,"temperature":1.0,"reasoning_tokens":1874,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T16:18:39.848311+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A reader could count every channel that contained a cell, record why each was excluded, and recompute the division-time distribution and the added-size versus birth-size regression with the excluded channels included. If the peak moves above 22 h or the slope moves toward zero, the reported sizer strategy is an artifact of channel selection.","supporting_citations":[{"cited_title":"A size-dependent division strategy accounts for leukemia cell size heterogeneity.Communications Physics, 7(1), July 2024","cited_arxiv_id":null,"evidence_quote":"Gives the indirect flow-cytometry estimate of a size-dependent division strategy that the paper's slope $m=-0.24$ directly confirms."},{"cited_title":"Cellpose: a generalist algorithm for cellular segmentation.Nature methods, 18(1):100–106, 2021","cited_arxiv_id":null,"evidence_quote":"Provides the deep-learning segmentation method used to obtain cell masks and lineage statistics."},{"cited_title":"Effect of a dual inlet channel on cell loading in microfluidics.Biomi- crofluidics, 8(6), 2014","cited_arxiv_id":null,"evidence_quote":"Supports the dual-inlet configuration that prevents cell accumulation near the loading port."},{"cited_title":"Size control in mammalian cells involves modulation of both growth rate and cell cycle duration","cited_arxiv_id":null,"evidence_quote":"Underpins the definitions of birth and division sizes used to compute added size and division symmetry."},{"cited_title":"Optical volume and mass measurements show that mammalian cells swell during mitosis.Journal of Cell Biology, 211(4):765–774, 2015","cited_arxiv_id":null,"evidence_quote":"Supports the timing convention for measuring size before cytokinesis."},{"cited_title":"Analysis of factors limiting bacterial growth in pdms mother machine devices.Fron- tiers in microbiology, 9:871, 2018","cited_arxiv_id":null,"evidence_quote":"Documents how PDMS mother-machine devices prolong division times, the baseline the paper compares its 22 h peak against."},{"cited_title":"Development and application of a cultiva- tion platform for mammalian suspension cell lines with single-cell resolution.Biotechnology and bioengineering, 118(2):992–1005, 2021","cited_arxiv_id":null,"evidence_quote":"Provides a prior cultivation platform for mammalian suspension cells with single-cell resolution, the comparison point for this device."}],"review_version":1}