{"id":"732ccb4f-f6f8-48cb-b068-1794fe293d3e","arxiv_id":"2601.12455","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":17,"one_line_summary":"A literature-parameterized ODE model of five TNBC cell populations ranks tumor proliferation, CAF-driven growth support, and M2-macrophage support as the biggest drivers of Day-168 tumor burden.","lead":"This paper builds a five-equation mathematical model of the triple-negative breast cancer tumor microenvironment and ranks which parameters drive tumor size at 24 weeks using global sensitivity analysis. The top-ranked parameters point to tumor proliferation rate and support from cancer-associated fibroblasts and M2 macrophages, consistent with already-known candidate targets.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"CAF/M2 influence in the sensitivity ranking is largely an artifact of the arbitrary β=4×initial-value rule and uniform ±50% ranges, not a robust therapeutic signal.","rationale":"The reader's CONDITIONAL verdict already captures the core fragility: parameter representativeness. My stress-test agrees and sharpens it to a specific, testable mechanism. The β parameters are set by a construction rule (4× initial value), and the sensitivity analysis then treats these constructed values as if they were empirical, applying the same ±50% range to all parameters. This makes threshold parameters like β_g artificially influential, because the initial condition sits at a fixed fraction of the threshold. The ranking's top entries are dominated by tumor-intrinsic or direct proliferation-boost parameters (α_j, α_g, β_g, K_B, α_f), not by the immune-suppression pathways highlighted in the Discussion. Without calibration to clinical TNBC dynamics or an alternative parameter-uncertainty scheme, the 'most-influential parameters' cannot be separated from the modeling convention used to set them. The proposed test—changing the β multiplier and rerunning the Sobol analysis—directly probes whether the ranking is robust. I therefore recommend no change to the reader's CONDITIONAL verdict; the work remains a useful hypothesis-generating framework but not a validated identification of therapeutic targets.","tokens_in":40730,"tokens_out":7576,"duration_ms":87725,"concrete_test":"Recompute total-order Sobol indices after resetting all β parameters to 10× the corresponding initial values (i.e., β = k·initial with k=10), keeping initial values and all other parameters at their original nominal values and using the same ±50% uniform ranges and 175,000 base samples. If β_g and the other threshold parameters leave the top five, or the top CAF/M2 pathway assignments change, the original ranking is an artifact of the 4× rule rather than a robust model property.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim—that CAF/M2 parameters disproportionately control tumor burden—rests on the Sobol ranking computed with nominal β thresholds set by the rule 'take the corresponding cell population's initial value and multiply it by 4' (Parameter Calculations; rows 2, 4, 7, 15, 17, 20, 22, 24, 26, 30, 32, 36, 38 of Table 3). β is therefore a pure modeling convention, not a measured quantity, and the construction places every initial population at exactly 20% of the Michaelis-Menten saturation point. With uniform ±50% ranges, perturbing β around this operating point is guaranteed to produce a non-negligible Sobol index. The #3 ranked parameter is β_g (threshold for CAF boost on tumor proliferation), derived solely from 4×722=2888 mm^3. If the multiplier were different, the sensitivity of β and the 'disproportionate influence' of the CAF/M2 pathways would change. The paper provides no calibration or external validation that could separate this structural artifact from biological importance. Additionally, the Discussion labels the top parameters as 'CAF and M2 macrophage-mediated immune suppression,' but α_g and α_f actually represent direct boosts to tumor proliferation, not immune suppression, so the interpretive leap is not supported by the model's own outputs.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a five-population ODE model of the TNBC tumor microenvironment (M2 macrophages, CAFs, tumor cells, CTLs, and Tregs) and performs a global Sobol sensitivity analysis to identify parameters that most influence tumor volume at day 168. The authors report that the top-ranked parameters correspond to CAF- and M2-related pathways and interpret this as supporting the relevance of stroma- and immune-mediated tumor support as therapeutic targets. The model is intended as a hypothesis-generation tool; treatment simulations are not included.","tokens_in":41202,"tokens_out":2494,"duration_ms":32840,"significance":"If the sensitivity ranking were robust to parameter uncertainty, the paper would provide a useful quantitative framework for prioritizing combination-therapy hypotheses in TNBC. The Sobol computation itself is well powered (175,000 base samples, 7,175,000 model evaluations) and follows standard Saltelli estimators. The authors are transparent about many parameter estimates being drawn from non-TNBC settings or marked as 'estimated.' However, the central claim that CAF/M2 pathways disproportionately control tumor burden is not yet supported, because the ranking is heavily influenced by an arbitrary beta-parameter rule and the model is not calibrated or validated. The strongest contribution is the model structure and the transparent parameter table; the weakest part is the leap from sensitivity indices to therapeutic-target conclusions without robustness checks or external validation.","major_comments":[{"comment":"All beta parameters are set as 4 times the corresponding initial population, e.g., beta_g = 4 x 722 = 2888 mm^3 (row 17). This arbitrary rule fixes every Michaelis-Menten term at 20% saturation at baseline. With uniform +/-50% ranges, the Sobol indices for these betas are essentially guaranteed to be non-negligible. In particular, beta_g (#3) and beta_f (#15) are not biologically measured; their high ranking is an artifact of the chosen multiplier. The authors should test robustness by varying the multiplier (e.g., 2, 6, 10) or by using literature-based saturation estimates, and show whether the ranking of CAF/M2 parameters persists.","section":"Parameter Calculations / Beta Parameters; Table 3"},{"comment":"The model is never calibrated or validated against observed TNBC tumor dynamics. The discussion claims the sensitivity analysis identifies parameters that 'exert a disproportionate influence on tumor burden,' but the nominal parameters and ranges are not grounded in TNBC-specific longitudinal data. Many parameters come from melanoma, lung cancer, LCMV-infected mice, or chick-embryo fibroblasts. Without calibration or a predictive check against independent data, the sensitivity ranking cannot support therapeutic-target conclusions. At minimum, the authors should perform a model-to-data comparison (e.g., patient tumor growth trajectories or xenograft data) or clearly reframe the claims as purely structural sensitivity of an unvalidated model.","section":"Model and Justifications / Simulations; Eq. (3)"},{"comment":"The interpretation that the top parameters correspond to 'CAF and M2 macrophage-mediated immune suppression' is not supported by the model. In Eq. (3), alpha_g and alpha_f directly multiply the logistic tumor growth term as boosts to proliferation, not immune-suppression pathways. The top-ranked parameters include the intrinsic proliferation rate and carrying capacity, followed by alpha_g, beta_g, and alpha_f, which are direct proliferative modulators. The sensitivity analysis does not separate immune suppression from direct stromal/macrophage support; the Discussion's phrasing overstates what the model identifies.","section":"Discussion; Figure 4; Eq. (3)"},{"comment":"The choice of uniform +/-50% ranges for all parameters is not justified, and the sensitivity indices are known to depend on the input distributions. The paper does not report any robustness analysis over alternative distribution shapes or widths. For parameters with strong biological constraints (e.g., death rates, thresholds with units inconsistent in Table 3), uniform ranges may place significant mass in implausible regions. The authors should either justify the ranges with data or include a distributional sensitivity check to demonstrate that the top-ranking parameters are stable.","section":"Simulations and Sensitivity Analysis; uniform distributions"}],"minor_comments":[{"comment":"The caption says 'Each mechanism is labeled and discussed in detail in section blank.' The cross-reference is missing; please fill in the appropriate section.","section":"Figure 1 caption"},{"comment":"The units for the CTL-related threshold parameters (beta_h and beta_t) are listed as 'cells/mm^3*day'; they should be 'cells/mm^3' to match the Michaelis-Menten form used in the equations.","section":"Table 3, rows 7 and 20"},{"comment":"The text describes converting a percentage of tumor volume to a number of Tregs, then states 'We adjusted this value to 10.9e3 based on our clinical authors’ input.' The preceding computed number is 8.91e7 cells/mm^3, so the adjustment is a large reduction. Please clarify the reasoning and the exact final value used in Table 2.","section":"Initial Value Calculations / Tregs"},{"comment":"The caption says the blue curve is produced by 'freezing the top 5' most-influential parameters, while the text earlier says the top six parameters are used for the orange curve. Please make the statement consistent.","section":"Figure 6 caption"},{"comment":"The equations are not explicitly displayed in the text; the descriptions refer to pathways (a, b, c, etc.) but the ODE system itself is not written out in the manuscript. For reproducibility, the full equations should be included in the main text or in a clearly numbered appendix.","section":"Model and Justifications / Equations (1)–(5)"},{"comment":"Several 'estimated' parameters (e.g., s_M2 = 10,000 cells/mm^3/day, d_m = 0.000998/day) are presented without a derivation or justification beyond 'clinical expertise.' A supplementary table with a brief rationale or a sensitivity analysis excluding these parameters would help readers judge their influence.","section":"General"}],"recommendation":"major_revision","confidential_remarks":"The manuscript has a useful model construction and a transparent, well-powered sensitivity analysis, but the central conclusions currently overreach the evidence. The arbitrary beta rule and the lack of calibration/validation are fixable with additional robustness analyses and reframing; I therefore recommend major revision rather than rejection. Please ensure the authors explicitly connect the sensitivity ranking to the model equations and avoid claiming 'immune suppression' for parameters that directly promote proliferation."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"What you should know: this paper builds a five-population ODE for the TNBC tumor microenvironment and runs a Sobol sensitivity analysis on 39 parameters. It is clearly written, transparent about data sources, and the sensitivity computation is competently done. But the headline conclusion—that CAF and M2 pathways are the most influential therapeutic targets—is not as robust as the abstract implies. The top ranking is dominated by tumor intrinsic parameters plus the CAF/M2 boost terms that were parameterized from studies already showing these cells help tumors grow. And the #3 parameter, β_g, is set by an arbitrary rule: 4× the initial CAF volume. That is a modeling convention, not a measurement.\n\nThe specific combination of five populations with this interaction structure is new, and the parameter table is a useful compilation. The authors take care to distinguish measured from estimated values. The Sobol run is well powered (175k samples, 7.1M evaluations), and the pairwise plots are a nice touch. As a hypothesis-generation tool for thinking about combination therapy, it is a reasonable starting point.\n\nThe main problem is that the sensitivity ranking is only meaningful if nominal parameters and ranges represent real Stage II TNBC. Many values come from melanoma, lung cancer, LCMV-infected mice, or chick-embryo fibroblasts; thresholds are labeled “estimated”; and the β=4×-initial-value rule sets every population at 20% of saturation. With uniform ±50% ranges, this essentially guarantees non-negligible Sobol indices for those thresholds. So the specific ranking—β_g at #3—is partly an artifact. The absence of calibration or validation against observed TNBC dynamics means there is no way to separate structural convention from biological importance. Also, the discussion calls these “CAF and M2 immune suppression,” but α_g and α_f are direct boosts to tumor proliferation. That is an overreach.\n\nThe paper is honest about its limitations, so I do not think it is misleading on purpose. It is a scaffold for future work, not a validated target list. For a reader building TME models or teaching sensitivity analysis, it is useful; for a reader looking for robust therapeutic targets, it is not there yet.\n\nShould it go to peer review? Yes—it is coherent, well-scoped, and the field needs more careful sensitivity analyses. But the reviewers should push for a sensitivity analysis over the beta rule itself (e.g., different multipliers) and for at least one external validation check before claiming target identification.","headline":"A clearly-built but under-validated ODE sensitivity framework; the top target ranking is real but partly a consequence of arbitrary threshold values.","tokens_in":41693,"tokens_out":2153,"would_cite":false,"duration_ms":26606,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["92C50","92C42","34A34"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper claims that a global sensitivity analysis of a five-population ODE model identifies the CAF-driven boost to tumor proliferation and M2-macrophage-driven immune suppression as the most influential regulators of TNBC tumor burden ov","keywords":["triple-negative breast cancer","tumor microenvironment","mathematical model","ordinary differential equations","global sensitivity analysis","Sobol indices","cancer-associated fibroblasts","M2 macrophages"],"falsifier":"Re-estimate every beta parameter from TNBC-specific measurements (replacing the 'initial value times four' rule) and re-run the Sobol analysis; if the top-six ranking changes, the paper's central ordering is an artifact of its estimation rule. Alternatively, measure the CAF fold-boost (alpha_g) in human TNBC co-cultures: the sensitivity analysis samples only 0.5 to 1.5 times the nominal value of 1.63, so a measured value near 1 would demote CAF support from its current rank.","tokens_in":40598,"feed_emoji":"🧬","tokens_out":5937,"duration_ms":64409,"temperature":0.7,"pith_summary":"This paper tries to establish that a simple five-population ordinary differential equation model of the triple-negative breast cancer tumor microenvironment, when subjected to a global variance-based sensitivity analysis, reveals a small set of parameters that dominate tumor burden over the 168-day treatment-relevant window. The authors argue that, after the expected tumor-intrinsic parameters, the most influential pathways are those by which cancer-associated fibroblasts and M2 macrophages boost tumor growth and suppress cytotoxic T-cell activity. If true, the model gives a quantitative, literature-grounded tool for prioritizing combination-therapy hypotheses in a cancer with few targeted options. The paper is careful to present the findings as hypothesis-generating rather than predictions of therapeutic efficacy.","feed_headline":"Sensitivity test names CAF/M2 pathways as TNBC tumor drivers","feed_subtitle":"Global sensitivity analysis of a five-cell ODE model points combination therapy at stromal support.","key_machinery":"The central machinery is the five-equation ODE model together with the Sobol variance-based global sensitivity analysis. Michaelis-Menten terms keep all boosting and inhibiting interactions bounded for biological plausibility, and the Sobol total-order indices rank each parameter's combined direct and interaction-driven contribution to the day-168 tumor volume. The histogram and pairwise-projection analyses are then used to show that varying only the top five or six parameters reproduces most of the variability generated by varying all thirty-nine, and to visualize interaction effects between the leading parameters.","core_discovery":"The authors construct a system of five coupled ODEs for M2 macrophages, cancer-associated fibroblasts, TNBC tumor cells, activated cytotoxic T lymphocytes, and regulatory T cells, with each interaction term bounded by Michaelis-Menten kinetics. They then run a Sobol global sensitivity analysis over 39 parameters, taking the computed tumor volume at day 168 as the quantity of interest. Their central claim is that the resulting ranking identifies a small subset of parameters that exert a disproportionate influence on tumor burden, and that the most-influential parameters correspond to CAF-mediated and M2-macrophage-mediated support of tumor growth — mechanisms consistent with emerging therapeu","pith_inferences":["A natural extension is to re-estimate every beta parameter from TNBC-specific measurements rather than the 'initial value times four' rule, and then check whether the top-six ranking persists.","Because the quantity of interest is day-168 tumor volume, the ranking favours pathways that act on early growth; a long-term or recurrence endpoint may shift priorities toward immune-regulatory parameters.","The aggregation of M1 macrophages and CD4+ T cells into implicit effects means the 'M2' and 'CAF' pathways may summarize several distinct molecular mechanisms, so the hits should be resolved at the level of specific factors (e.g., Chi3L1, TGF-beta, CXCL12) before clinical planning.","The same sensitivity-analysis protocol could be applied to other aggressive cancers with similar stromal-immune landscapes to see whether CAF/M2 dominance is a general phenomenon or specific to TNBC."],"forward_implications":["The top influencers are the TNBC proliferation rate and tumor carrying capacity, so any therapy that slows intrinsic growth should still be the primary lever.","CAF-driven tumor proliferation and M2-macrophage-driven tumor proliferation are the most influential microenvironment parameters, indicating stromal support as a high-priority target.","The pairwise analysis shows that a fast-proliferating tumor can still be kept small by a weak CAF boost, suggesting CAF-secreted factors as a combination-therapy component.","Parameters should be read as net effects of several biological mechanisms, so the ranking points to pathway-level intervention points rather than specific molecules.","The framework supports future in silico testing of combination regimens, including optimal-control approaches."],"fun_headline_variants":["Math model pinpoints CAF/M2 support as TNBC drivers","Sensitivity analysis flags CAF/M2 pathways in TNBC","Mathematical model IDs stromal drivers in triple-negative breast cancer","Computational model reveals key TNBC tumor support mechanisms","Model ranks CAF/M2 interactions as top TNBC therapeutic targets"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The sensitivity ranking is meaningful only if the nominal parameter values and their uncertainty ranges represent human Stage II TNBC; many values are estimated, taken from other cancers or species, or set by the rule that beta equals the corresponding initial cell count times four.","fun_headline_variants_meta":{"raw":{"variants":["Math model pinpoints CAF/M2 support as TNBC drivers","Sensitivity analysis flags CAF/M2 pathways in TNBC","Mathematical model IDs stromal drivers in triple-negative breast cancer","Computational model reveals key TNBC tumor support mechanisms","Model ranks CAF/M2 interactions as top TNBC therapeutic targets"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000856,"raw_usage":{"total_tokens":3532,"prompt_tokens":698,"completion_tokens":2834,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":442,"completion_tokens_details":{"reasoning_tokens":2749}},"tokens_in":442,"tokens_out":2834,"duration_ms":20243,"temperature":1.0,"reasoning_tokens":2749,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T09:46:03.584037+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-estimate every beta parameter from TNBC-specific measurements (replacing the 'initial value times four' rule) and re-run the Sobol analysis; if the top-six ranking changes, the paper's central ordering is an artifact of its estimation rule. Alternatively, measure the CAF fold-boost (alpha_g) in human TNBC co-cultures: the sensitivity analysis samples only 0.5 to 1.5 times the nominal value of 1.63, so a measured value near 1 would demote CAF support from its current rank.","supporting_citations":[],"review_version":1}