pith. sign in

arxiv: 1907.10284 · v1 · pith:QKZ55G7Pnew · submitted 2019-07-24 · 🧬 q-bio.QM

Inferring long-range interactions between immune and tumor cells -- pitfalls and (partial) solutions

Claus Metzner This is my paper

Pith reviewed 2026-05-24 16:48 UTC · model grok-4.3

classification 🧬 q-bio.QM
keywords immune cellstumor cellscell trajectoriesinteraction inferenceturning anglesmaximum likelihoodchemotaxiscell migration
0
0 comments X

The pith

A turning angle probability model correctly infers cell-cell interactions even when migration behavior switches over time.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper develops ways to tell whether immune cells are actively drawn to tumor cells over long distances by analyzing their recorded paths in a dish. A model that pulls cells toward targets with a distance-dependent force recovers the right parameters when cells keep steady migration habits, but it reports fake attractions when cells independently alter their speed or direction over time. An alternative model that instead adjusts the probability of turning toward or away from the nearest tumor cell based on distance recovers the true interaction parameters in both steady and switching cases. This distinction matters for experiments because real cells frequently change behavior without external signals, which can mimic directed interactions if not modeled carefully.

Core claim

The distance-dependent turning angle probability model recovers the correct interaction parameters even with temporally switching cell migration, while the force model detects spurious interactions in independent cells that change migration behavior.

What carries the argument

Distance-dependent turning angle probability model, which fits the probabilities of positive and negative turning angles of an immune cell as a function of its distance to the nearest tumor cell via maximum likelihood.

Load-bearing premise

Simulated trajectories with constant or temporally switching migration properties adequately represent the confounding factors and data characteristics present in actual experimental cell recordings.

What would settle it

Generate trajectories of non-interacting cells programmed to switch migration directions or speeds at random times, then verify whether the turning angle model returns near-zero interaction strength while the force model returns positive strength.

Figures

Figures reproduced from arXiv: 1907.10284 by Claus Metzner.

Figure 1
Figure 1. Figure 1: FIG. 1 [PITH_FULL_IMAGE:figures/full_fig_p015_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2 [PITH_FULL_IMAGE:figures/full_fig_p016_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3 [PITH_FULL_IMAGE:figures/full_fig_p017_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4 [PITH_FULL_IMAGE:figures/full_fig_p018_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5 [PITH_FULL_IMAGE:figures/full_fig_p019_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: FIG. 6 [PITH_FULL_IMAGE:figures/full_fig_p020_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: FIG. 7 [PITH_FULL_IMAGE:figures/full_fig_p021_7.png] view at source ↗
read the original abstract

Upcoming immunotherapies for cancer treatment rely on the ability of the immune system to detect and eliminate tumors in the body. A highly simplified version of this process can be studied in a Petri dish: starting with a random distribution of immune and tumor cells, it can be observed in detail how individual immune cells migrate towards nearby tumor cells, establish contact, and attack. Nevertheless, it remains unclear whether the immune cells find their targets by chance, or if they approach them 'on purpose', using remote sensing mechanisms such as chemotaxis. In this work, we present methods to infer the strength and range of long-range cell-cell interactions from time-lapse recorded cell trajectories, using a maximum likelihood method to fit the model parameters. First, we model the interactions as a distance-dependent 'force' that attracts immune cells towards their nearest tumor cell. While this approach correctly recovers the interaction parameters of simulated cells with constant migration properties, it detects spurious interactions in the case of independent cells that spontaneously change their migration behavior over time. We therefore use an alternative approach that models the interactions by distance-dependent probabilities for positive and negative turning angles of the migrating immune cell. We demonstrate that the latter approach finds the correct interaction parameters even with temporally switching cell migration.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 0 minor

Summary. The manuscript claims that a distance-dependent force model for inferring long-range immune-tumor cell interactions from trajectories recovers true parameters in simulations with constant migration properties but produces spurious interactions when cells independently switch migration behavior over time; an alternative model using distance-dependent probabilities for positive/negative turning angles recovers the correct parameters even under temporal switching, as demonstrated via maximum-likelihood fitting on forward-simulated trajectories with known ground truth.

Significance. If the turning-angle approach holds under more realistic conditions, the work identifies a substantive pitfall in force-based inference methods commonly used for cell trajectory analysis and provides a concrete alternative. Credit is due for the use of simulations with independently set interaction parameters and behavioral switches, which supplies an external (non-circular) check on recovery rather than reducing to fitted quantities by construction.

major comments (1)
  1. [Simulation validation (abstract and results)] The central claim—that the turning-angle probability model recovers true interaction parameters under temporally switching migration while the force model yields false positives—rests on the assumption that the simulated trajectories adequately capture the confounding factors in real experimental cell recordings. No quantitative match is shown between simulated and experimental turning-angle distributions, speed histograms, or persistence statistics (see abstract and simulation results).

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their constructive review and recommendation. Our response to the single major comment is provided below. The manuscript uses controlled simulations with known ground truth to demonstrate a methodological pitfall and a proposed solution; we defend that this does not require quantitative matching to experimental distributions.

read point-by-point responses

  1. Referee: [Simulation validation (abstract and results)] The central claim—that the turning-angle probability model recovers true interaction parameters under temporally switching migration while the force model yields false positives—rests on the assumption that the simulated trajectories adequately capture the confounding factors in real experimental cell recordings. No quantitative match is shown between simulated and experimental turning-angle distributions, speed histograms, or persistence statistics (see abstract and simulation results).

    Authors: We respectfully disagree that the central claim requires quantitative matching between simulated and experimental statistics. The manuscript's core contribution is a controlled demonstration, via forward simulations with independently specified ground-truth interaction parameters and behavioral switches, that the force-based model produces spurious detections while the turning-angle probability model recovers the true parameters. This isolates temporal heterogeneity as a confounding factor without circularity. The simulations incorporate persistence, speed variation, and distance-dependent turning to capture the relevant mechanism, but the validation is external (recovery of known parameters) rather than a claim to reproduce any specific experimental dataset. The abstract and results sections make no assertion that the simulated trajectories match real recordings; the focus is methodological robustness. Adding such matches would require new experimental data and is outside the current scope, though we would be willing to expand the discussion of simulation parameterization if helpful. revision: no

Circularity Check

0 steps flagged

No significant circularity; validation uses independent forward simulations

full rationale

The paper's derivation chain consists of defining two alternative interaction models (force-based and turning-angle probability) and testing parameter recovery on simulated trajectories. Interaction parameters are set independently in the simulations before fitting, and recovery success is measured against those known values. This constitutes an external check rather than any reduction of outputs to inputs by construction. No self-definitional equations, fitted-input predictions, or load-bearing self-citations appear in the provided text. The central claim therefore remains independent of the fitting procedure itself.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 0 invented entities

The claim rests on maximum likelihood recovery from trajectories under two alternative interaction models; free parameters are the strength and range of distance dependence in each model, with the domain assumption that turning statistics or forces fully capture interaction effects.

free parameters (2)
  • interaction strength
    Magnitude parameter of the distance-dependent force or turning probability, fitted by maximum likelihood to trajectories.
  • interaction range
    Distance scale parameter governing how far the force or turning bias extends, fitted by maximum likelihood.
axioms (2)
  • standard math Maximum likelihood estimation recovers true parameters when data are generated from the assumed interaction model.
    Invoked to justify fitting both the force and turning models to simulated trajectories.
  • domain assumption Cell migration can be decomposed into either distance-dependent forces or distance-dependent turning probabilities plus independent behavioral switches.
    Core modeling choice separating the two approaches tested in the abstract.

pith-pipeline@v0.9.0 · 5747 in / 1205 out tokens · 27417 ms · 2026-05-24T16:48:47.002240+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

13 extracted references · 13 canonical work pages · 2 internal anchors

  1. [1]

    Immunobiology: the immune system in health and disease , volume 7

    Charles A Janeway, Paul Travers, Mark Walport, Mark Shlomchik, et al. Immunobiology: the immune system in health and disease , volume 7. Current Biology London, 1996. 12

    work page 1996

  2. [2]

    Pathogen recognition by the innate immune system

    Himanshu Kumar, Taro Kawai, and Shizuo Akira. Pathogen recognition by the innate immune system. International reviews of immunology , 30(1):16–34, 2011

    work page 2011

  3. [3]

    Cancer immunotherapy

    Manfred Schuster, Andreas Nechansky, and Ralf Kircheis. Cancer immunotherapy. Biotech- nology Journal: Healthcare Nutrition Technology , 1(2):138–147, 2006

    work page 2006

  4. [4]

    Decade in reviewcancer immunotherapy: entering the mainstream of cancer treatment

    Steven A Rosenberg. Decade in reviewcancer immunotherapy: entering the mainstream of cancer treatment. Nature Reviews Clinical Oncology, 11(11):630, 2014

    work page 2014

  5. [5]

    Breathing new life into immunother- apy: review of melanoma, lung and kidney cancer

    Charles G Drake, Evan J Lipson, and Julie R Brahmer. Breathing new life into immunother- apy: review of melanoma, lung and kidney cancer. Nature reviews Clinical oncology, 11(1):24, 2014

    work page 2014

  6. [6]

    Distinct migration and contact dynamics of resting and il-2-activated human natural killer cells

    Per Erik Olofsson, Elin Forslund, Bruno Vanherberghen, Ksenia Chechet, Oscar Mickelin, Alexander Rivera Ahlin, Tobias Everhorn, and Bj¨ orn¨Onfelt. Distinct migration and contact dynamics of resting and il-2-activated human natural killer cells. Frontiers in immunology , 5:80, 2014

    work page 2014

  7. [7]

    Bystander cells enhance nk cytotoxic efficiency by reducing search time

    Xiao Zhou, Renping Zhao, Karsten Schwarz, Matthieu Mangeat, Eva C Schwarz, Mohamed Hamed, Ivan Bogeski, Volkhard Helms, Heiko Rieger, and Bin Qu. Bystander cells enhance nk cytotoxic efficiency by reducing search time. Scientific reports, 7:44357, 2017

    work page 2017

  8. [8]

    Chemotaxis

    Michael Eisenbach. Chemotaxis. World Scientific Publishing Company, 2004

    work page 2004

  9. [9]

    Random walk models in biology

    Edward A Codling, Michael J Plank, and Simon Benhamou. Random walk models in biology. Journal of the Royal society interface , 5(25):813–834, 2008

    work page 2008

  10. [10]

    Analytical prediction of the spatiotemporal distribution of chemoattractants around their source: Theory and application to complement-mediated chemotaxis

    Volkmar Heinrich, Wooten D Simpson III, and Emmet A Francis. Analytical prediction of the spatiotemporal distribution of chemoattractants around their source: Theory and application to complement-mediated chemotaxis. Frontiers in immunology, 8:578, 2017

    work page 2017

  11. [11]

    Superstatistical analysis and modelling of heterogeneous random walks

    Claus Metzner, Christoph Mark, Julian Steinwachs, Lena Lautscham, Franz Stadler, and Ben Fabry. Superstatistical analysis and modelling of heterogeneous random walks. Nature communications, 6(May):7516, jun 2015

    work page 2015

  12. [12]

    Principles of efficient chemotactic pursuit

    Claus Metzner. Principles of efficient chemotactic pursuit. arXiv:1902.10589, 2019

    work page internal anchor Pith review Pith/arXiv arXiv 1902

  13. [13]

    Detecting long-range attraction between migrating cells based on p-value distributions

    Claus Metzner. Detecting long-range attraction between migrating cells based on p-value distributions. arXiv:1906.08481, 2019. 13 ADDITIONAL INFORMA TION Author contributions statement: CM has devised the study, developed the simulation programs, performed and interpreted the simulations, and wrote the paper. Funding: This work was funded by the Grant ME1...

    work page internal anchor Pith review Pith/arXiv arXiv 1906