Reproducible Analysis of Cross-Site Distribution Shift in High-Dimensional Microbiome Classifiers
Fuente:
Zenodo
Salvato in:
| Autore principale: | |
|---|---|
| Natura: | Recurso digital |
| Pubblicazione: |
Zenodo
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866901923460808704 |
|---|---|
| author | Pereira, Antonio |
| author_facet | Pereira, Antonio |
| contents | <p>This repository contains all scripts, harmonized feature matrices, and derived outputs required to reproduce the analyses presented in:</p> <p>“Site-Level Distribution Shift Dominates Disease Signal and Undermines Cross-Site Generalization in High-Dimensional Microbiome Classifiers.”</p> <p>The study examines cross-site generalization under distribution shift in high-dimensional biomedical classifiers, using three independent Parkinson’s disease microbiome cohorts (Finland, Malaysia, USA; n = 682 subjects) as a case study. Analyses include stratified within-cohort validation, leave-one-cohort-out evaluation, cross-cohort transfer assessment, calibration degradation analysis, and feature stability quantification.</p> <h3>Extended Description</h3> <p>This repository provides:</p> <ul> <li> <p>Harmonized genus-level count matrix (209 shared genera)</p> </li> <li> <p>Cohort metadata used for modeling</p> </li> <li> <p>Elastic net modeling scripts (R 4.4.2)</p> </li> <li> <p>Stratified cross-validation implementation</p> </li> <li> <p>Leave-one-cohort-out (LOCO) portability analysis</p> </li> <li> <p>Cross-cohort transfer matrices</p> </li> <li> <p>Permutation-based null controls</p> </li> <li> <p>Calibration analysis (Brier score)</p> </li> <li> <p>Variance partitioning via PERMANOVA</p> </li> <li> <p>Feature stability metrics (Jaccard overlap, effective sparsity, sign concordance)</p> </li> <li> <p>All CSV outputs used to generate manuscript tables and figures</p> </li> </ul> <p>The deposited processed matrices correspond exactly to the feature space used for all modeling and statistical analyses reported in the manuscript. Running the primary analysis script reproduces all reported performance values.</p> |
| format | Recurso digital |
| id | zenodo_https___doi_org_10_5281_zenodo_18674302 |
| institution | Zenodo |
| language | |
| publishDate | 2026 |
| publisher | Zenodo |
| record_format | zenodo |
| spellingShingle | Reproducible Analysis of Cross-Site Distribution Shift in High-Dimensional Microbiome Classifiers Pereira, Antonio distribution shift cross-site validation model generalization biomedical machine learning high-dimensional data elastic net <p>This repository contains all scripts, harmonized feature matrices, and derived outputs required to reproduce the analyses presented in:</p> <p>“Site-Level Distribution Shift Dominates Disease Signal and Undermines Cross-Site Generalization in High-Dimensional Microbiome Classifiers.”</p> <p>The study examines cross-site generalization under distribution shift in high-dimensional biomedical classifiers, using three independent Parkinson’s disease microbiome cohorts (Finland, Malaysia, USA; n = 682 subjects) as a case study. Analyses include stratified within-cohort validation, leave-one-cohort-out evaluation, cross-cohort transfer assessment, calibration degradation analysis, and feature stability quantification.</p> <h3>Extended Description</h3> <p>This repository provides:</p> <ul> <li> <p>Harmonized genus-level count matrix (209 shared genera)</p> </li> <li> <p>Cohort metadata used for modeling</p> </li> <li> <p>Elastic net modeling scripts (R 4.4.2)</p> </li> <li> <p>Stratified cross-validation implementation</p> </li> <li> <p>Leave-one-cohort-out (LOCO) portability analysis</p> </li> <li> <p>Cross-cohort transfer matrices</p> </li> <li> <p>Permutation-based null controls</p> </li> <li> <p>Calibration analysis (Brier score)</p> </li> <li> <p>Variance partitioning via PERMANOVA</p> </li> <li> <p>Feature stability metrics (Jaccard overlap, effective sparsity, sign concordance)</p> </li> <li> <p>All CSV outputs used to generate manuscript tables and figures</p> </li> </ul> <p>The deposited processed matrices correspond exactly to the feature space used for all modeling and statistical analyses reported in the manuscript. Running the primary analysis script reproduces all reported performance values.</p> |
| title | Reproducible Analysis of Cross-Site Distribution Shift in High-Dimensional Microbiome Classifiers |
| topic | distribution shift cross-site validation model generalization biomedical machine learning high-dimensional data elastic net |
| url | https://doi.org/10.5281/zenodo.18674302 |