pith. sign in

arxiv: 2606.24351 · v1 · pith:GT7AGUKVnew · submitted 2026-06-23 · 🧬 q-bio.NC · physics.data-an

Graph-based analysis of inflammatory profiles in New Onset Refractory Status Epilepticus (NORSE)

Linon Denis , Martin Guillemaud , Vincent Navarro , Aur\'elie Hanin , Mario Chavez This is my paper

Pith reviewed 2026-06-25 21:56 UTC · model grok-4.3

classification 🧬 q-bio.NC physics.data-an
keywords cNORSEcytokine profilinggraph clusteringinflammatory clustersstatus epilepticusattribution modelimmune heterogeneity
0
0 comments X

The pith

Graph clustering of 96-cytokine profiles from 62 cNORSE patients creates stable inflammatory groups that new cases can be assigned to with probability and confidence scores.

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

The paper builds a graph-clustering method on cytokine data from 62 patients with cryptogenic new-onset refractory status epilepticus to define distinct inflammatory groups. It then develops attribution models that, for any new cytokine profile, output the most likely group, an attribution probability, and a statistical confidence measure derived from Monte-Carlo simulations and custom parametric tests. The resulting clinician interface processes profiles quickly and can track changes in inflammatory state over time. A sympathetic reader would care because cNORSE shows marked heterogeneity in immune signatures that currently hinders selection of targeted immunotherapies.

Core claim

Building on graph clustering that identified biologically validated inflammatory groups in a 62-patient cohort with 96-cytokine serum profiles, the authors construct attribution models that assign new profiles to these groups while supplying probability and confidence values; Monte-Carlo simulations and parametric tests quantify the reliability of each assignment, and the framework is packaged as a clinician-friendly interface that supports both single-timepoint classification and longitudinal trajectory tracking.

What carries the argument

Graph clustering on 96-cytokine profiles followed by tailored attribution models that compute group membership probability and statistical confidence via Monte-Carlo simulations and custom parametric tests.

If this is right

  • Clinicians receive a most-likely inflammatory cluster, attribution probability, and confidence score for each new cytokine panel within seconds.
  • Longitudinal cytokine measurements can be mapped onto the same clusters to monitor how a patient's inflammatory state evolves.
  • The framework supplies a concrete route toward matching individual patients to specific immunomodulatory therapies based on cluster membership.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If the clusters prove reproducible across centers, the same attribution machinery could be applied to other cryptogenic inflammatory encephalopathies that also show cytokine heterogeneity.
  • The method supplies a ready-made stratification variable that future clinical trials of immunotherapy in status epilepticus could use to enrich for likely responders.
  • Because the output includes explicit confidence values, the approach naturally supports a 'watch-and-wait' strategy for borderline cases rather than immediate treatment escalation.

Load-bearing premise

The inflammatory clusters found by graph clustering in the original 62-patient cohort are biologically distinct and stable enough that new, unseen profiles can be reliably assigned to them.

What would settle it

An independent cohort of cNORSE patients whose cytokine profiles are run through the attribution model yields low probabilities or confidence scores for every cluster, or the clusters themselves change when the original cohort is re-clustered after adding new data.

Figures

Figures reproduced from arXiv: 2606.24351 by Aur\'elie Hanin, Linon Denis, Mario Chavez, Martin Guillemaud, Vincent Navarro.

Figure 1
Figure 1. Figure 1: Incorporation of a new patient to the reference patient graph. From the adja [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The random simulation procedure for adding a new patient node (represented in [PITH_FULL_IMAGE:figures/full_fig_p010_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The interface with the proposed pipeline. From left to right: data upload [PITH_FULL_IMAGE:figures/full_fig_p014_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Evolution of the attribution probability distributions per cluster in the SP [PITH_FULL_IMAGE:figures/full_fig_p015_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Attribution probability distributions in the SP model per cluster, per patient [PITH_FULL_IMAGE:figures/full_fig_p018_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Attribution probabilities in the SP model per cluster, per patient of the NORS [PITH_FULL_IMAGE:figures/full_fig_p019_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Inflammatory profiling results for patient P82 assigned to cluster B with an attri [PITH_FULL_IMAGE:figures/full_fig_p022_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Summary of cluster transitions (arrows) observed in the longitudinal study on [PITH_FULL_IMAGE:figures/full_fig_p023_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Extended reference cohort study. A. Cluster assignment with the extended cohort of 82 NORSE patients. B. Clustered graph structure with the extended cohort. repeated the clustering procedure on an extended reference cohort compris￾ing both NORSEREF and NORSENEW patients with complete 96-cytokine panels (N=82 NORSE patients, compared to N=62 in the original cohort). As shown in [PITH_FULL_IMAGE:figures/ful… view at source ↗
read the original abstract

Background and Objectives: Cryptogenic new-onset refractory status epilepticus (cNORSE) represents one of the most severe forms of status epilepticus, occurring in patients without prior neurological disease, and remaining of unknown aetiology despite extensive diagnostic evaluation. Emerging evidence supports a role for immune dysregulation in cNORSE; however, marked heterogeneity in inflammatory signatures has been reported, complicating the selection of targeted immunotherapies. Therefore, a critical need for tools facilitating the interpretation of cytokine panels exists. Methods: Building on the identification of distinct inflammatory groups of cNORSE patients using a graph clustering approach applied to a cohort of 62 patients with serum profiling of 96 cytokines, we tailored new models to quantify attribution probability to biologically validated clusters. Statistical assessment of the most informative model involved Monte-Carlo simulations and custom-developed parametric tests. Ultimately, we applied our framework to the implementation of a clinician-friendly interface for inflammatory profiling. Results: Our approach enables quick processing of several cytokine profiles, providing the most likely inflammatory cluster, associated attribution probability, and statistical confidence. For longitudinal assessments, the proposed method may also allow tracking the evolution of inflammatory trajectories over time. Conclusion: Systematic statistical characterization of the inflammatory heterogeneity in cNORSE requires the development of clinically actionable support tools. Our study offers a framework that may support personalized immunomodulatory strategies in cNORSE patients through clustering-based cytokine profiling.

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

2 major / 1 minor

Summary. The manuscript applies graph clustering to serum profiles of 96 cytokines from a 62-patient cNORSE cohort to identify inflammatory subgroups, develops attribution models that assign new profiles to these clusters together with probability and statistical-confidence estimates, validates the models via Monte-Carlo simulations and custom parametric tests, and implements a clinician-facing interface for rapid profiling and longitudinal tracking.

Significance. If the clusters are shown to be stable and the attribution generalizable, the framework could supply a concrete, statistically grounded tool for interpreting the marked inflammatory heterogeneity in cNORSE and thereby support individualized immunomodulatory decisions. The explicit use of Monte-Carlo simulations and parametric tests for the attribution step is a methodological strength that supplies quantifiable performance measures conditional on the discovered partition.

major comments (2)
  1. [Methods (graph clustering subsection)] Methods (graph clustering subsection): no cluster-stability diagnostics (bootstrap resampling, adjusted Rand index across resamples, or leave-one-out reproducibility) are reported for the 62-patient, 96-cytokine graph partition. With n=62 in a high-dimensional space, the absence of such checks directly undermines the assumption that the clusters are biologically fixed rather than sample-specific, which is load-bearing for any claim of reliable out-of-sample attribution.
  2. [Results (statistical assessment of attribution model)] Results (statistical assessment of attribution model): the Monte-Carlo simulations and parametric tests evaluate model performance conditional on the fixed clusters obtained from the full cohort; they do not test whether the underlying partition itself reappears under resampling. Consequently the reported attribution probabilities and confidence intervals for unseen profiles rest on an unverified premise.
minor comments (1)
  1. [Abstract] Abstract: the phrase 'biologically validated clusters' is used without stating the external criteria or independent data employed for that validation.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the detailed and constructive review. The two major comments both concern the lack of explicit cluster-stability assessment. We agree that this is a substantive methodological gap for claims of generalizable attribution and will add the requested diagnostics in revision.

read point-by-point responses

  1. Referee: [Methods (graph clustering subsection)] Methods (graph clustering subsection): no cluster-stability diagnostics (bootstrap resampling, adjusted Rand index across resamples, or leave-one-out reproducibility) are reported for the 62-patient, 96-cytokine graph partition. With n=62 in a high-dimensional space, the absence of such checks directly undermines the assumption that the clusters are biologically fixed rather than sample-specific, which is load-bearing for any claim of reliable out-of-sample attribution.

    Authors: We acknowledge that the original Methods section did not include formal cluster-stability diagnostics. Although the graph construction and Louvain partitioning followed standard practice, the referee is correct that stability under resampling is required to support the downstream attribution claims. In the revised manuscript we will add a dedicated subsection describing (i) 1000 bootstrap resamples of the 62-patient cohort, (ii) computation of adjusted Rand indices between each resampled partition and the original partition, and (iii) leave-one-out reproducibility of cluster membership. These results, together with a stability summary table, will be reported in a new Results subsection. revision: yes

  2. Referee: [Results (statistical assessment of attribution model)] Results (statistical assessment of attribution model): the Monte-Carlo simulations and parametric tests evaluate model performance conditional on the fixed clusters obtained from the full cohort; they do not test whether the underlying partition itself reappears under resampling. Consequently the reported attribution probabilities and confidence intervals for unseen profiles rest on an unverified premise.

    Authors: The Monte-Carlo and parametric tests were intentionally conditioned on the observed partition because the immediate clinical goal was to assign new cytokine profiles to the clusters discovered in the present cohort. Nevertheless, the referee correctly notes that this leaves the stability of the partition itself untested. The bootstrap stability analysis described in response to the first comment will be used to quantify how often the same partition structure re-emerges; attribution probabilities and confidence intervals will then be reported both conditionally on the original partition and as averages across stable resamples, with the difference discussed explicitly. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper performs graph clustering on the 62-patient, 96-cytokine cohort to discover inflammatory groups, then constructs separate attribution models whose performance is evaluated via Monte-Carlo simulations and parametric tests. These steps are data-driven and externally assessed; the attribution probabilities are not shown to be equivalent to the clustering inputs by construction, nor do any load-bearing claims reduce to self-citations or fitted parameters renamed as predictions. The derivation chain is therefore self-contained against the described benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract supplies insufficient detail to enumerate free parameters, axioms, or invented entities.

pith-pipeline@v0.9.1-grok · 5794 in / 956 out tokens · 27935 ms · 2026-06-25T21:56:33.544873+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

30 extracted references

  1. [1]

    Trinka, H

    E. Trinka, H. Cock, D. Hesdorffer, A. O. Rossetti, I. E. Scheffer, S. Shin- nar, S. Shorvon, D. H. Lowenstein, A definition and classification of status epilepticus–report of the ilae task force on classification of status epilepticus, Epilepsia 56 (10) (2015) 1515–1523

    2015

  2. [2]

    L. J. Hirsch, N. Gaspard, A. van Baalen, R. Nabbout, S. Demeret, T. Loddenkemper, V. Navarro, N. Specchio, L. Lagae, A. O. Rossetti, et al., Proposed consensus definitions for new-onset refractory status epilepticus (norse), febrile infection-related epilepsy syndrome (fires), and related conditions, Epilepsia 59 (4) (2018) 739–744

    2018

  3. [3]

    Périn, W

    B. Périn, W. Szurhaj, New onset refractory status epilepticus: state of the art, Revue Neurologique 178 (1-2) (2022) 74–83

    2022

  4. [4]

    Carson, C

    R. Carson, C. M. Stredny, Severe, refractory seizures: new-onset refrac- tory status epilepticus and febrile infection-related epilepsy syndrome, Medical Clinics 108 (1) (2024) 201–213

    2024

  5. [5]

    Tharmaraja, J

    T. Tharmaraja, J. S. Y. Ho, A. Neligan, S. Rajakulendran, The etiol- ogy and mortality of new-onset refractory status epilepticus (norse) in 31 adults: a systematic review and meta-analysis, Epilepsia 64 (5) (2023) 1113–1124

    2023

  6. [6]

    Wickström, O

    R. Wickström, O. Taraschenko, R. Dilena, E. T. Payne, N. Specchio, R. Nabbout, S. Koh, N. Gaspard, L. J. Hirsch, I. N. C. Group, et al., International consensus recommendations for management of new onset refractory status epilepticus (norse) including febrile infection-related epilepsy syndrome (fires): summary and clinical tools, Epilepsia 63 (11) (2022...

    2022

  7. [7]

    Hanin, E

    A. Hanin, E. Muscal, L. J. Hirsch, Second-line immunotherapy in new onset refractory status epilepticus, Epilepsia 65 (5) (2024) 1203–1223

    2024

  8. [8]

    Hanin, J

    A. Hanin, J. Cespedes, K. Dorgham, Y. Pulluru, M. Gopaul, G. Goro- chov, D. A. Hafler, V. Navarro, N. Gaspard, L. J. Hirsch, Cytokines in new-onset refractory status epilepticus predict outcomes, Annals of Neurology 94 (1) (2023) 75–90

    2023

  9. [9]

    Guillemaud, M

    M. Guillemaud, M. Chavez, F. Kobeissy, A. Vezzani, A. D. Jimenez, M. M. Basha, A. Batra, S. Demeret, O. Eka, K. Eschbach, et al., Iden- tification of distinct biological groups of patients with cryptogenic norse via inflammatory profiling, Neurology: Neuroimmunology & Neuroin- flammation 12 (4) (2025) e200403

    2025

  10. [10]

    S. F. Seys, H. Scheers, P. Van den Brande, G. Marijsse, E. Dilissen, A. Van Den Bergh, P. C. Goeminne, P. W. Hellings, J. L. Ceuppens, L. J. Dupont, et al., Cluster analysis of sputum cytokine-high profiles 32 reveals diversity in t (h) 2-high asthma patients, Respiratory Research 18 (1) (2017) 39

    2017

  11. [11]

    Y. Wang, R. B. Meagher, S. Ambati, H. Cheng, P. Ma, B. G. Phillips, Patients with obstructive sleep apnea have altered levels of four cy- tokines associated with cardiovascular and kidney disease, but near nor- mal levels with airways therapy, Nature and Science of Sleep (2021) 457–466

    2021

  12. [12]

    Bañales-Méndez, M

    L.Barrera, E.Montes-Servín, A.Barrera, L.Ramírez-Tirado, F.Salinas- Parra, J. Bañales-Méndez, M. Sandoval-Ríos, Ó. Arrieta, Cytokine pro- file determined by data-mining analysis set into clusters of non-small-cell lung cancer patients according to prognosis, Annals of Oncology 26 (2) (2015) 428–435

    2015

  13. [13]

    J. H. Turner, R. K. Chandra, P. Li, K. Bonnet, D. G. Schlundt, Identifi- cation of clinically relevant chronic rhinosinusitis endotypes using cluster analysis of mucus cytokines, Journal of Allergy and Clinical Immunology 141 (5) (2018) 1895–1897

    2018

  14. [14]

    J. Bai, C. Wu, D. Zhong, D. Xu, Q. Wang, X. Zeng, Hierarchical clus- ter analysis of cytokine profiles reveals a cutaneous vasculitis-associated subgroup in dermatomyositis, Clinical Rheumatology 40 (3) (2021) 999– 1008

    2021

  15. [15]

    Tomassen, G

    P. Tomassen, G. Vandeplas, T. Van Zele, L.-O. Cardell, J. Arebro, H. Olze, U. Förster-Ruhrmann, M. L. Kowalski, A. Olszewska-Ziąber, G. Holtappels, et al., Inflammatory endotypes of chronic rhinosinusitis 33 based on cluster analysis of biomarkers, Journal of Allergy and Clinical Immunology 137 (5) (2016) 1449–1456

    2016

  16. [16]

    J. C. Morse, M. H. Shilts, K. A. Ely, P. Li, Q. Sheng, L.-C. Huang, T. J. Wannemuehler, N. I. Chowdhury, R. K. Chandra, S. R. Das, et al., Patterns of olfactory dysfunction in chronic rhinosinusitis identified by hierarchical cluster analysis and machine learning algorithms, Interna- tional Forum of Allergy & Rhinology 9 (3) (2019) 255–264

    2019

  17. [17]

    V. D. Blondel, J.-L. Guillaume, R. Lambiotte, E. Lefebvre, Fast unfold- ing of communities in large networks, Journal of statistical mechanics: theory and experiment 2008 (10) (2008) P10008

    2008

  18. [18]

    Girvan, M

    M. Girvan, M. E. Newman, Community structure in social and biologi- cal networks, Proceedings of the National Academy of Sciences 99 (12) (2002) 7821–7826

    2002

  19. [19]

    M. E. Newman, M. Girvan, Finding and evaluating community structure in networks, Physical Review E 69 (2) (2004) 026113

    2004

  20. [20]

    P. Pons, M. Latapy, Computing communities in large networks using random walks, Journal of Graph Algorithms and Applications 10 (2) (2006) 191–218

    2006

  21. [21]

    M. E. Newman, Finding community structure in networks using the eigenvectors of matrices, Physical Review E 74 (3) (2006) 036104

    2006

  22. [22]

    S. S. Meena, P. C. Sharma, Y. P. Singh, M. P. Singh, Graph embedding based label propagation for community detection in social networks, Scientific Reports 15 (1) (2025) 41817. 34

    2025

  23. [23]

    R. C. Wang, Z. Wang, Precision medicine: disease subtyping and tai- lored treatment, Cancers 15 (15) (2023) 3837

    2023

  24. [24]

    Lambert, A.-L

    J. Lambert, A.-L. Leutenegger, A.-S. Jannot, A. Baudot, Tracking clus- ters of patients over time enables extracting information from medico- administrative databases, Journal of Biomedical Informatics 139 (2023) 104309

    2023

  25. [25]

    Greene, D

    D. Greene, D. Doyle, P. Cunningham, Tracking the evolution of com- munities in dynamic social networks, in: 2010 international conference on advances in social networks analysis and mining, IEEE, 2010, pp. 176–183

    2010

  26. [26]

    Irani, N

    J. Irani, N. Pise, M. Phatak, Clustering techniques and the similarity measures used in clustering: A survey, International Journal of Com- puter Applications 134 (7) (2016) 9–14

    2016

  27. [27]

    Hanin, J

    A. Hanin, J. Cespedes, Y. Pulluru, M. Gopaul, E. Aronica, D. De- campo, I. Helbig, C. L. Howe, A. Huttner, S. Koh, et al., Review and standard operating procedures for collection of biospecimens and anal- ysis of biomarkers in new onset refractory status epilepticus, Epilepsia 64 (6) (2023) 1444–1457

    2023

  28. [28]

    D. J. Polley, P. Latham, M. Y. Choi, K. A. Buhler, M. J. Fritzler, M. L. Fritzler, Identification of novel clusters of co-expressing cytokines in a diagnostic cytokine multiplex test, Frontiers in Immunology 14 (2023) 1223817. 35

    2023

  29. [29]

    Kothur, S

    K. Kothur, S. Bandodkar, L. Wienholt, S. Chu, A. Pope, D. Gill, R. C. Dale, Etiology is the key determinant of neuroinflammation in epilepsy: elevation of cerebrospinal fluid cytokines and chemokines in febrile infection-related epilepsy syndrome and febrile status epilepti- cus, Epilepsia 60 (8) (2019) 1678–1688

    2019

  30. [30]

    Fang, et al., Circulating cytokine profiling and clustering identify biomarker predicting efficacy of ici in combination with chemotherapy, Cancer Letters 631 (2025) 217918

    X.Xu, Y.Li, S.Sun, X.Lin, W.Zhang, Y.Wu, B.Wei, D.Xu, C.Zheng, H. Fang, et al., Circulating cytokine profiling and clustering identify biomarker predicting efficacy of ici in combination with chemotherapy, Cancer Letters 631 (2025) 217918. 36 Appendix A. Supplementary Methods The CN model To attribute clusters to a newly added patient to graph eG=G∪(V N+1...

    2025