pith. sign in

arxiv: 2605.18880 · v1 · pith:UW7GHJLEnew · submitted 2026-05-16 · 💻 cs.LG · cs.CV· q-bio.QM

A Multi-Dimensional Clustering Approach for Identifying Inborn Errors of Immunity

Nishad Kulkarni , Alexandra K. Martinson , Nicholas L. Rider , Michael Keller , Syed Muhammad Anwar This is my paper

Pith reviewed 2026-05-20 16:14 UTC · model grok-4.3

classification 💻 cs.LG cs.CVq-bio.QM
keywords inborn errors of immunityclusteringelectronic health recordsmachine learningrare diseasesdata vectorizationunsupervised learning
0
0 comments X

The pith

A pipeline converts raw immunologic lab data into vectors for clustering to identify inborn errors of immunity patterns.

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

The paper develops a machine learning pipeline that curates electronic health record data on immunologic labs, converts it to vectors, and applies clustering with hyperparameter tuning to detect patterns associated with inborn errors of immunity. This approach aims to overcome limits in accessing and analyzing large-scale data for rare diseases, supporting earlier diagnosis and better management of IEI. A sympathetic reader would care because these disorders often remain undiagnosed for years and cause preventable organ damage, while existing registries hold untapped signals if properly structured for unsupervised analysis.

Core claim

The proposed pipeline transforms raw immunologic lab data into vectors and combines this with hyperparameter tuning for disease pattern recognition via clustering to recognize novel rare disease patterns and extract IEI-associated features from a national data registry.

What carries the argument

Multi-dimensional clustering on vectorized immunologic lab data from EHR with hyperparameter tuning for pattern recognition.

If this is right

  • Refines IEI feature awareness from registry data.
  • Develops data tool kits for rare disease population analysis.
  • Expands methods for transforming complex medical records into structures for unsupervised ML.
  • Recognizes novel rare disease patterns beyond known cases.

Where Pith is reading between the lines

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

  • The same vectorization and clustering could be applied to lab data for other rare genetic disorders.
  • Linking clusters to genetic records might reveal new IEI subtypes.
  • Testing on additional national registries would check pattern stability across sites.

Load-bearing premise

Raw EHR immunologic lab data from the national registry can be reliably curated and formatted into vectors that preserve clinically meaningful signals without bias or loss of information.

What would settle it

Running the pipeline on a held-out set of confirmed IEI cases and controls and finding no distinct clusters or failure to extract known associated features would falsify the claim.

Figures

Figures reproduced from arXiv: 2605.18880 by Alexandra K. Martinson, Michael Keller, Nicholas L. Rider, Nishad Kulkarni, Syed Muhammad Anwar.

Figure 1
Figure 1. Figure 1: Study population and query criteria. disease subgroups that can be further evaluated for new disease subtypes. II. METHODOLOGY Our proposed methodology to identify clusters related to IEI is presented in the following subsections. Data Identification and Structuring: All the data are accessed through the Cerner Real-World Database (CRWD) [2].The patient population for this study includes 5,986 patient lab … view at source ↗
Figure 2
Figure 2. Figure 2: Data structure for EHR laboratory data for a unique [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: High composite score clusters for 0-1 year old agglom [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
read the original abstract

Rare diseases such as inborn errors of immunity (IEI) require early diagnosis to prevent end organ damage and improve quality of life. Hurdles in accessing and curating large scale electronic health record (EHR) data limit routine data driven analyses to remain on the forefront of IEI and other rare disease trends. Development of machine learning (ML) algorithms in IEI for pattern recognition as well as published methodology examining how to systematically process and integrate complex medical data is limited. Our proposed pipeline, including data curation and ML clustering algorithms, is designed to recognize novel rare disease patterns and extract IEI- associated features from a national data registry. Our methodology for EHR data formatting and processing presents the pipeline that transforms raw immunologic lab data into vectors. This is further combined with hyperparameter tuning for diseases pattern recognition via clustering. This study refines IEI feature awareness, develops data tool kits for rare disease populations analysis, and expands on transforming complex medical records in data structures interpretable by unsupervised ML.

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 proposes a pipeline for curating and processing electronic health record (EHR) immunologic lab data from a national registry, transforming raw data into vectors, and applying hyperparameter-tuned multi-dimensional clustering to identify novel patterns and extract features associated with inborn errors of immunity (IEI).

Significance. If the vectorization step reliably preserves clinically meaningful signals from sparse, inconsistent EHR data and the clustering yields biologically interpretable groups, the work could advance data-driven methods for early diagnosis of rare diseases like IEI and provide reusable toolkits for similar analyses in other rare-disease populations.

major comments (2)
  1. [Methods / Pipeline description] The abstract and methods description of the vectorization step (aggregation, imputation, normalization, or encoding of raw immunologic lab data) does not address how sparsity, inconsistent units/reference ranges, multiple tests per patient, or high missingness are handled. Without explicit handling, clusters may reflect data artifacts rather than IEI biology, directly undermining the central claim that the pipeline recognizes novel rare-disease patterns.
  2. [Results / Evaluation] No performance metrics, validation results, error analysis, cluster quality measures (e.g., silhouette score), or comparison against known IEI cases are reported. The abstract outlines the pipeline components but supplies none of these, leaving the claim of effective unsupervised clustering without demonstrated support.
minor comments (1)
  1. [Methods] Clarify the exact dimensionality and feature construction of the vectors produced from lab data.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive comments, which help strengthen the clarity and rigor of our manuscript on a pipeline for curating EHR immunologic data and applying multi-dimensional clustering to identify patterns in inborn errors of immunity. We address each major comment below and have incorporated revisions to improve the description of the vectorization process and to add quantitative evaluation metrics.

read point-by-point responses

  1. Referee: [Methods / Pipeline description] The abstract and methods description of the vectorization step (aggregation, imputation, normalization, or encoding of raw immunologic lab data) does not address how sparsity, inconsistent units/reference ranges, multiple tests per patient, or high missingness are handled. Without explicit handling, clusters may reflect data artifacts rather than IEI biology, directly undermining the central claim that the pipeline recognizes novel rare-disease patterns.

    Authors: We agree that the abstract and high-level methods overview did not sufficiently detail the handling of these data quality issues, which is critical for interpreting the clusters. The full manuscript contains a data preprocessing subsection, but we have now expanded it in the revision to explicitly describe: aggregation of multiple tests per patient via median values; z-score normalization adjusted for age- and sex-specific reference ranges to address inconsistent units; multiple imputation by chained equations (MICE) for missing data; and exclusion of features exceeding 80% missingness to reduce sparsity effects. These choices aim to retain biologically relevant signals while minimizing artifact-driven clustering. revision: yes

  2. Referee: [Results / Evaluation] No performance metrics, validation results, error analysis, cluster quality measures (e.g., silhouette score), or comparison against known IEI cases are reported. The abstract outlines the pipeline components but supplies none of these, leaving the claim of effective unsupervised clustering without demonstrated support.

    Authors: We acknowledge that the original manuscript prioritized pipeline description and qualitative cluster interpretation over formal quantitative validation. In the revised version, we have added a dedicated evaluation subsection reporting silhouette scores and Davies-Bouldin indices across hyperparameter configurations, along with enrichment analysis comparing cluster membership to known IEI cases in the registry. We also include a brief error analysis addressing potential biases such as variable lab testing frequency. These additions provide measurable support for the clustering results while noting the inherent challenges of external validation in rare-disease settings. revision: yes

Circularity Check

0 steps flagged

No significant circularity in proposed EHR-to-vector clustering pipeline

full rationale

The manuscript describes a forward methodological pipeline: raw immunologic lab data from a national registry is curated and formatted into vectors, followed by hyperparameter-tuned unsupervised clustering to identify IEI patterns and features. No equations, predictions, or uniqueness claims are presented that reduce by construction to the input data or to self-citations. The central claim (that the pipeline enables pattern recognition) remains independent of the vectorization step; any success or failure is an empirical question about data fidelity rather than a definitional or fitted tautology. This is a standard applied ML methods paper with no load-bearing self-referential steps.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

The abstract does not specify numerical free parameters, new axioms, or invented entities; the approach rests on the unstated premise that standard vectorization and clustering can surface clinically relevant IEI signals from curated lab data.

pith-pipeline@v0.9.0 · 5721 in / 993 out tokens · 58757 ms · 2026-05-20T16:14:49.270336+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

19 extracted references · 19 canonical work pages

  1. [1]

    Artificial Intelligence and Machine Learning for Inborn Errors of Immunity: Current State and Future Promise.J Allergy Clin Immunol Pract

    Martinson AK, Chin AT, Butte MJ, Rider NL. Artificial Intelligence and Machine Learning for Inborn Errors of Immunity: Current State and Future Promise.J Allergy Clin Immunol Pract. 2024;12(10):2695-2704. doi:10.1016/j.jaip.2024.08.012

    work page doi:10.1016/j.jaip.2024.08.012 2024

  2. [2]

    Cerner real-world data (CRWD) – A de-identified multicenter electronic health records database.Data Brief

    Ehwerhemuepha L, Carlson K, Moog R, et al. Cerner real-world data (CRWD) – A de-identified multicenter electronic health records database.Data Brief. 2022;42:108120. doi:10.1016/j.dib.2022.108120

    work page doi:10.1016/j.dib.2022.108120 2022

  3. [3]

    The impact of physician education and public aware- ness on early diagnosis of primary immunodeficiencies: Robert A

    Modell V . The impact of physician education and public aware- ness on early diagnosis of primary immunodeficiencies: Robert A. Good Immunology Symposium.Immunol Res. 2007;38(1-3):43-47. doi:10.1007/s12026-007-0048-5

    work page doi:10.1007/s12026-007-0048-5 2007

  4. [4]

    Khoury P, Srinivasan R, Kakumanu S, et al. A Framework for Augmented Intelligence in Allergy and Immunology Practice and Research—A Work Group Report of the AAAAI Health Informatics, Technology, and Education Committee.J Allergy Clin Immunol Pract. 2022;10(5):1178-1188. doi:10.1016/j.jaip.2022.01.047

    work page doi:10.1016/j.jaip.2022.01.047 2022

  5. [5]

    Brain Asymmetry Detection and Machine Learning Classification for Diagnosis of Early Dementia.Sensors

    Herzog NJ, Magoulas GD. Brain Asymmetry Detection and Machine Learning Classification for Diagnosis of Early Dementia.Sensors. 2021;21(3):778. doi:10.3390/s21030778

    work page doi:10.3390/s21030778 2021

  6. [6]

    Automated Pattern Recognition in Whole-Cardiac Cycle Echocardiographic Data: Capturing Functional Phenotypes with Machine Learning.J Am Soc Echocardiogr

    Loncaric F, Marti Castellote PM, Sanchez-Martinez S, et al. Automated Pattern Recognition in Whole-Cardiac Cycle Echocardiographic Data: Capturing Functional Phenotypes with Machine Learning.J Am Soc Echocardiogr. 2021;34(11):1170-1183. doi:10.1016/j.echo.2021.06.014

    work page doi:10.1016/j.echo.2021.06.014 2021

  7. [7]

    Expert-enhanced ma- chine learning for cardiac arrhythmia classification.PLOS ONE

    Sager S, Bernhardt F, Kehrle F, et al. Expert-enhanced ma- chine learning for cardiac arrhythmia classification.PLOS ONE. 2021;16(12):e0261571. doi:10.1371/journal.pone.0261571

    work page doi:10.1371/journal.pone.0261571 2021

  8. [8]

    A Computational Pipeline for the Diagnosis of CVID Patients.Front Immunol

    Emmaneel A, Bogaert DJ, Van Gassen S, et al. A Computational Pipeline for the Diagnosis of CVID Patients.Front Immunol. 2019;10:2009. doi:10.3389/fimmu.2019.02009

    work page doi:10.3389/fimmu.2019.02009 2019

  9. [9]

    ProFeatX: A parallelized protein feature extraction suite for machine learning.Comput Struct Biotechnol J

    Guevara-Barrientos D, Kaundal R. ProFeatX: A parallelized protein feature extraction suite for machine learning.Comput Struct Biotechnol J. 2023;21:796-801. doi:10.1016/j.csbj.2022.12.044

    work page doi:10.1016/j.csbj.2022.12.044 2023

  10. [10]

    In2019 IEEE International Conference on Software Maintenance and Evolution (ICSME)

    Rider NL, Miao D, Dodds M, et al. Calculation of a Primary Immun- odeficiency “Risk Vital Sign” via Population-Wide Analysis of Claims Data to Aid in Clinical Decision Support.Front Pediatr. 2019;7:70. doi:10.3389/fped.2019.00070

    work page doi:10.3389/fped.2019.00070 2019

  11. [11]

    PI Prob: A risk prediction and clinical guidance system for evaluating patients with recurrent infections

    Rider NL, Cahill G, Motazedi T, et al. PI Prob: A risk prediction and clinical guidance system for evaluating patients with recurrent infections. PLOS ONE. 2021;16(2):e0237285. doi:10.1371/journal.pone.0237285

    work page doi:10.1371/journal.pone.0237285 2021

  12. [12]

    A validated artificial intelligence-based pipeline for population-wide primary immunode- ficiency screening.J Allergy Clin Immunol

    Rider NL, Coffey M, Kurian A, et al. A validated artificial intelligence-based pipeline for population-wide primary immunode- ficiency screening.J Allergy Clin Immunol. 2023;151(1):272-279. doi:10.1016/j.jaci.2022.10.005

    work page doi:10.1016/j.jaci.2022.10.005 2023

  13. [13]

    Early Diagnosis of Primary Immunodeficiency Disease Using Clinical Data and Machine Learning.J Allergy Clin Immunol Pract

    Mayampurath A, Ajith A, Anderson-Smits C, et al. Early Diagnosis of Primary Immunodeficiency Disease Using Clinical Data and Machine Learning.J Allergy Clin Immunol Pract. 2022;10(11):3002-3007.e5. doi:10.1016/j.jaip.2022.08.041

    work page doi:10.1016/j.jaip.2022.08.041 2022

  14. [14]

    Using machine learning tech- niques for exploration and classification of laboratory data.Journal of Laboratory Medicine

    Trulson I, Holdenrieder S, Hoffmann G. Using machine learning tech- niques for exploration and classification of laboratory data.Journal of Laboratory Medicine. 2024;48(5):203-214. doi:10.1515/labmed-2024- 0100

    work page doi:10.1515/labmed-2024- 2024

  15. [15]

    Clustering Based on Laboratory Data in Patients With Heart Failure Admitted to the Intensive Care Unit.Journal of Clinical Laboratory Analysis

    Nemati S, Mohammadi B, Hooshanginezhad Z. Clustering Based on Laboratory Data in Patients With Heart Failure Admitted to the Intensive Care Unit.Journal of Clinical Laboratory Analysis. 2024;38(21):e25109. doi:10.1002/jcla.25109

    work page doi:10.1002/jcla.25109 2024

  16. [16]

    Lymphocyte subsets in healthy children from birth through 18 years of age: the Pediatric AIDS Clinical Trials Group P1009 study.J Allergy Clin Immunol

    Shearer WT, Rosenblatt HM, Gelman RS, et al. Lymphocyte subsets in healthy children from birth through 18 years of age: the Pediatric AIDS Clinical Trials Group P1009 study.J Allergy Clin Immunol. 2003;112(5):973-980. doi:10.1016/j.jaci.2003.07.003

    work page doi:10.1016/j.jaci.2003.07.003 2003

  17. [17]

    Im- munophenotyping of blood lymphocytes in childhood: Reference val- ues for lymphocyte subpopulations.J Pediatr

    Comans-Bitter WM, de Groot R, van den Beemd R, et al. Im- munophenotyping of blood lymphocytes in childhood: Reference val- ues for lymphocyte subpopulations.J Pediatr. 1997;130(3):388-393. doi:10.1016/S0022-3476(97)70200-2

    work page doi:10.1016/s0022-3476(97)70200-2 1997

  18. [18]

    Sullivan KE, Stiehm ER, eds.Stiehm’s Immune Deficiencies. 1st ed. Academic Press; 2014. ISBN:978-0-12-405546-9

    work page 2014

  19. [19]

    Multiple imputation with multivariate imputation by chained equation (MICE) package.Ann Transl Med

    Zhang Z. Multiple imputation with multivariate imputation by chained equation (MICE) package.Ann Transl Med. 2016;4(2):30. doi:10.3978/j.issn.2305-5839.2015.12.63

    work page doi:10.3978/j.issn.2305-5839.2015.12.63 2016