Corpus 1,272 assessed · 1,173 scored · 643 reproduced ≥75 · 168 flagged ·∅ 74.1/100
← New search

Methylation patterns of the nasal epigenome of hospitalized SARS-CoV-2 positive patients reveal insights into molecular mechanisms of COVID-19.

BMC Med Genomics · 2025
not yet assessed 2/4
Why this verdict

Part of the results reproduced; minor but material deviations remained.

Reproduced on the brainbox compute brainarbeit.com
✓ What held up
  • Nothing in this column.
What did not (or only partly)
  • 🟡Could not use the authors’ exact input data
  • 🔴Reported values were only indirectly comparable
  • 🔴A deviation arose in the data or preprocessing
  • 🔴A deviation was attributed to the published material
  • 🟡Reported values were not (fully) derivable from the shared data
  • 🟡The deviation was non-trivial in magnitude
  • 🟡The central claim did not (fully) hold under reproduction
  • 🟡Overall, the reproduction showed a material discrepancy
Reproduction agent’s raw note

DROP. The paper (nasal-epigenome WGBS in hospitalized COVID-19 patients) is described well at the methods level, but its reported numbers are NOT reproducible at 80/20. (1) Core caller is DRAGEN EP v2.6.3 (Illumina, proprietary FPGA hardware-locked) — not installable on «our HPC»/«infra» => env_unresolvable. (2) The authors' analysis code (methylKit tiling-window logistic regression + MethylSeekR v1.38) is NOT shipped; the RU's only code link is the third-party trimming tool Trim Galore (one early step) => no_code for every gradeable result. (3) The RU BRIEF's data accession GSE183071 is MIS-LINKED: it is the Gomez-Carballa external RNA/NanoString validation set (PMID 35202626, 'Expression profiling by array'), NOT this paper's methylation data; the paper's own WGBS data is BioProject PRJNA1162448 (positive, 61 Bisulfite-Seq runs ~45GB each) + GEO GSE168254 (negatives) — verified via NCBI/ENA. Per HARD RULE 2 (P16) I did NOT down-rank for third-party code; I scoped and PREPARED a faithful run of the named tool (Trim Galore v0.4.2 on real paper sample SRR30711030 = T-COV-R-306, bounded to 5M read pairs; see reproduction/run.sbatch) — but it was NOT executed because the «our HPC» VPN 2FA was not confirmed this session, and the paper reports no trim-stage statistic to grade against anyway. WHAT I DID NOT ATTEMPT: full WGBS alignment/methylation-calling (would require proprietary DRAGEN or a divergent Bismark substitute = the hard ~20% we are told not to chase), and all downstream methylKit/MethylSeekR/GREAT/HOMER/IPA numbers. FABRICATION: none observed at screening level; the data is genuine high-depth WGBS consistent with the Methods, but the reported numbers are UN-CHECKABLE (proprietary caller + unshipped scripts), not refuted. Artifacts: scope.md, original/claims.tsv, AUDIT.md, reproduction/agreement.json, reproduction/run.sbatch.

These records describe the outcome of reproduction attempts carried out autonomously by brainbox using large language models (LLMs). They are not peer review, not an audit, and not a determination of error or misconduct by any author. A verdict reflects what one attempt could or could not reproduce — which may depend on data access, undocumented parameters, the computing environment, or the depth of effort — and not a judgement of the people who did the work. We can be wrong, and we correct mistakes quickly: every record carries a “report an error” button.

Assessment versions

Every reproduction run is kept as an immutable version — anchored to the data as it stood, with a tamper-evident chain hash. A rerun (e.g. after an author updates a deposit) adds a new version; the previous one stays on record.

  1. v1 current initial assessment
    assessed: 2026-06-15 ⛓ 0825b56472df
✎ I am an author of this paper

Updated or fixed a deposit, or is there an erratum? Ask us to re-run the metrics. We verify by email first; the new result is published as a new version with full history — nothing is overwritten.

Reason for the rerun

We email you a confirmation link first. The rerun is an objective re-measurement — it cannot change the verdict in your favour, only ask us to look again.

Provenance — full disclosure

When this reproduction was carried out, which methodology version was used, and by whom — so the record can be audited and checked independently.

Reproduced
2026-06-15
Rubric version
v1.0
Assessed by
🤖 AI curator · claude (ai-curator room) · v1.0 · run #1 2026-06-15
no human curator yet
Last updated
2026-07-31

Provisional, curator- or AI-assessed, and independently checkable. A reproduction outcome states what one attempt could reproduce — not a judgement of the authors.

Deep full-text extraction

Model: opus
Founding hypothesis

Differential DNA methylation in the nasal mucosa (the target tissue of SARS-CoV-2) distinguishes severe from mild COVID-19 and reveals molecular drivers of disease severity, but it is unclear whether these methylation differences predispose to severe disease or arise in response to it.

Core claims
  • The nasal methylome shows differential DNA methylation in intergenic regions and low methylated regions (LMRs), highlighting distal regulatory/enhancer-like elements in COVID-19 gene regulation finding
  • Severe COVID-19 shows a preponderance of hypomethylated DMRs relative to hypermethylated DMRs in nasal samples finding
  • Differentially methylated pathways implicate immune cell recruitment/function and inflammatory signaling including PI3K/Akt, Notch, and NF-κB mechanism
  • Hypermethylation of the FUT4 (CD15) promoter implicates impaired neutrophil adhesion in severe disease finding
  • Hypermethylation of ELF5 binding sites suggests downregulation of ELF5 targets as a factor in COVID-19 phenotypic variability finding
  • WGBS of nasal mucosa can capture cell-type-specific methylation signatures and infer differential regulation of immune-mediated pathways method
  • COVID-19 positive individuals show greatly increased overlap of LMRs with immune (lymphoid and myeloid) cell regulatory elements versus COVID-19 negative individuals finding
  • DNA methylation serves as a marker of the host immune response to SARS-CoV-2 infection finding
Experimental setups
Assay System Perturbation Readout Platform
Whole genome bisulfite sequencing (WGBS) Nasal mucosa samples from SARS-CoV-2 positive subjects (4 severe/hospitalized, 57 mild/non-hospitalized) none (observational; SARS-CoV-2 infection severity) CpG percent methylation; UMRs/LMRs; differentially methylated regions (DMRs)
Whole genome bisulfite sequencing (WGBS) Nasal samples from COVID-19 negative / non-infected individuals (healthy controls, n=7 pooled) none UMR/LMR overlap with cell-type regulatory elements; promoter methylation
Methylation segmentation / regulatory region annotation (DHS overlap) Aggregated nasal WGBS datasets; DNase I hypersensitive sites across 16 cell types none UMR/LMR counts and overlap with cell-type-specific regulatory DNA
Differential methylation analysis (tiling window, logistic regression with age/race/gender covariates) Severe vs mild nasal WGBS datasets none Top DMRs ranked by q-value; hypo/hypermethylation; intergenic mapping
Pathway/GO enrichment analysis (Coronascape, GREAT, GO/KEGG/Reactome, IPA) Differentially methylated genes (DMGs) from severe vs mild cohorts none Enriched immune/inflammatory pathways; gene-pathway associations GREAT; Coronascape; QIAGEN IPA v01-21-03; UCSC Genome Browser
Key results
  • Hypermethylation of the FUT4 promoter in hospitalized vs non-hospitalized subjects (chr11:94,545,001–94,552,500) q = 2.90 × 10^-112
  • Severe COVID-19 had more hypomethylated DMRs (n=7,256) than hypermethylated DMRs (n=2,744) among top 10,000 DMRs 7,256 vs 2,744
  • COVID-19 positive vs negative: 46% vs 25% of LMRs overlapped immune cell regulatory elements 46% vs 25% (X2=5189.6)
  • COVID-19 positive vs negative: 24% vs 15% of LMRs overlapped lymphoid and 28% vs 12% overlapped myeloid regulatory elements lymphoid 24% vs 15%; myeloid 28% vs 12%
  • Relative hypomethylation of AKT1 promoter in severe vs mild disease methylation difference = -18.38%, q = 2.22 × 10^-22
  • Relative hypomethylation of ISG15 promoter (AKT1 downstream target) in severe vs mild disease methylation difference = -32.27%, q = 2.49 × 10^-35
  • Relative hypomethylation of ZEB2 and SNAI1 promoters in severe vs mild disease ZEB2 -24.39% (q=2.52×10^-32); SNAI1 -28.98% (q=6.11×10^-16)
  • Hypomethylated DMRs disproportionately in intergenic regions (49.6%) vs hypermethylated (13.8%) 49.6% vs 13.8% (X2=1064.4)
Key statistics
  • pvalue X2 = 5189.6, df = 1, p < 2.2 × 10^-16 (LMR overlap with immune regulatory elements, COVID+ vs COVID- (46% vs 25%))
  • pvalue X2 = 4557.1, df = 1, p < 2.2 × 10^-16 (Myeloid LMR overlap COVID+ vs COVID- (28% vs 12%))
  • pvalue X2 = 1264.0, df = 1, p < 2.2 × 10^-16 (Lymphoid LMR overlap COVID+ vs COVID- (24% vs 15%))
  • fold_change methylation difference = -32.27%, q = 2.49 × 10^-35 (ISG15 promoter hypomethylation severe vs mild)
  • fold_change q = 2.90 × 10^-112 (FUT4 promoter hypermethylation hospitalized vs non-hospitalized)
  • count 19,187 UMRs (avg 2,366 bp, ~117 CpGs); 43,924 LMRs (avg 642 bp, ~8 CpGs) (Regulatory regions identified across aggregated nasal WGBS samples)
  • count 13.2 million CpGs per sample at >10X coverage (Average WGBS sequencing depth)
  • pvalue X2 = 1,064.4, df = 1, p < 2.2 × 10^-16 (Intergenic enrichment of hypomethylated vs hypermethylated DMRs (49.6% vs 13.8%))

Statistical methods review

Model: sonnet

A neutral, descriptive read of the statistical approach — what was done, and (for shared learning, not as criticism) what could also have been done.

The study used whole-genome bisulfite sequencing (WGBS) of nasal samples from 61 SARS-CoV-2-positive individuals (4 severe/hospitalized, 57 mild/non-hospitalized) to identify differentially methylated regions (DMRs). DMR detection employed tiling-window logistic regression with age, race, and gender as covariates; the top 10,000 DMRs ranked by q-value were retained for downstream analysis. Proportion comparisons of regulatory-element overlaps between groups used chi-squared tests, and pathway enrichment was performed via Coronascape against GO, KEGG, and Reactome databases with p-value thresholds.

Replicationbiological Sample sizen=61 COVID-19 positive (4 severe/hospitalized, 57 mild/non-hospitalized); approximately 7 COVID-19 negative controls mentioned for UMR/LMR characterization; no formal power calculation described GroupsSevere (hospitalized, n=4) vs mild (non-hospitalized, n=57) COVID-19; COVID-19 positive (n=61) vs COVID-19 negative (~7) Pairingunpaired Randomization/blindingnot stated DispersionIQR Exact p-valuesyes Effect sizesyes Confidence intervalsno Multiplicity correctionq-value (FDR-based; specific algorithm not named)
Statistical tests used
Test Applied to n Assumptions
Logistic regression (tiling-window DMR analysis) Identification of DMRs between severe (n=4) and mild (n=57) COVID-19 nasal WGBS data; covariates: age, race, gender 61 (4 severe, 57 mild) not stated
Chi-squared test (Χ²) Comparison of proportions of LMRs and UMRs overlapping immune/lymphoid/myeloid/epithelial regulatory elements between COVID-19 positive and negative individuals null not stated
Pathway enrichment analysis (Coronascape / hypergeometric or Fisher's exact, method not specified) GO Biological Processes, KEGG Pathways, Reactome Gene Sets enrichment of hypomethylated and hypermethylated DMGs in severe vs mild COVID-19 487 hypomethylated genes; 503 hypermethylated genes not stated
GREAT genomic annotation (binomial / hypergeometric region-gene association) Gene annotation of hypo- and hypermethylated DMRs 10,000 top DMRs not stated
Approaches that could also have been used
  • The primary DMR comparison (severe vs mild) used logistic regression applied to groups of n=4 and n=57, a highly imbalanced and very small severe group.
    Could also: Permutation-based or exact tests (e.g., Fisher's exact test per CpG region, or bootstrap-based inference), or dedicated small-n WGBS tools such as DSS (which uses a beta-binomial model and empirical Bayes shrinkage) or MethylKit. — When one group is very small (n=4), asymptotic likelihood-based regression can have inflated Type I error and unstable covariate estimates; shrinkage estimators and permutation approaches borrow strength across the genome and can yield better-calibrated inference under extreme imbalance.
  • Chi-squared tests were used to compare proportions of UMRs/LMRs overlapping regulatory elements across multiple cell-type categories without multiplicity adjustment.
    Could also: A single omnibus chi-squared or Fisher's exact test on the full contingency table followed by post-hoc comparisons with Bonferroni or BH-FDR correction across cell types. — Testing multiple cell-type categories (immune, lymphoid, myeloid, epithelial, pulmonary) inflates the family-wise error rate; a correction such as Benjamini-Hochberg would control the false discovery rate across these related comparisons while remaining interpretable.
  • Pathway enrichment for DMGs used a p-value threshold (≤0.01 and ≤10⁻³) across hundreds of tested pathways without an explicit correction for the number of pathways evaluated.
    Could also: Benjamini-Hochberg FDR correction applied to all pathway p-values, with reporting of adjusted q-values alongside raw p-values. — Testing hundreds of GO/KEGG/Reactome pathways simultaneously inflates false positives; FDR adjustment is a standard practice in enrichment analysis and would clarify which pathway associations survive correction.
  • The study did not describe cell-type deconvolution of the bulk nasal WGBS signal; differences between groups were characterized by overlap with reference DHS maps.
    Could also: Reference-based epigenomic deconvolution (e.g., EpiDISH, MethylCIBERSORT, or BLUEPRINT-based cell-type reference panels) to estimate proportions of epithelial, immune, and stromal cell populations in each sample. — Bulk methylation differences between severe and mild disease could reflect changes in cell-type composition rather than (or in addition to) within-cell-type methylation changes; deconvolution would allow these contributions to be quantified and, if desired, statistically adjusted.
  • Effect sizes for DMRs were reported as raw methylation percentage differences (e.g., −18.38% for AKT1); the logistic regression framework could also yield odds ratios or standardized coefficients but these were not reported.
    Could also: Reporting of standardized effect sizes (Cohen's d on M-values, or odds ratios with 95% confidence intervals from the logistic regression) alongside methylation differences. — Raw methylation differences are intuitive but depend on the absolute methylation level; odds ratios or standardized measures facilitate comparison across loci with different background methylation levels and across studies.
  • Dispersion of the primary outcome (methylation) across subjects was not summarized at the regional level (e.g., no within-group SD or CI shown for DMRs); Table 1 demographics use IQR for age only.
    Could also: Reporting per-group mean ± SD (or median with IQR) of % methylation, or 95% confidence intervals around the methylation difference, for key DMRs highlighted in figures. — With n=4 severe cases, within-group variability is informative about the reliability of point estimates; CIs or SDs around the methylation difference would allow readers to gauge precision, especially given the small severe-group sample size.
Software: QIAGEN Ingenuity Pathway Analysis (IPA) 01-21-03 · Coronascape (pathway enrichment) · GREAT (Genomic Regions Enrichment of Annotations Tool) · UCSC Genome Browser · WGBS alignment/DMR pipeline (unspecified)

Citation network

Where this publication sits in the reproducibility-weighted citation graph — what it is built on, and what is built on it. Citation data from OpenAlex.

Citations
0
Impact: low
Foundation confidence
None of its references are in our reproducibility record yet — its foundation cannot be assessed.
Topics

No assessed neighbours yet — the network grows as more papers are assessed.

Data lineage

The datasets this paper uses (text-mined from the full text via Europe PMC), and which other assessed papers stand on the same data. A shared dataset is a factual link — not a judgement.

GSE183071 GEO in Methods (http://purl.org/orb/Methods)
no other assessed paper uses this yet
GSE212865 GEO in Methods (http://purl.org/orb/Methods)
no other assessed paper uses this yet
rs117126460 RefSNP in Results (http://purl.org/orb/Results)
no other assessed paper uses this yet

Downstream reach in the literature

8 downstream papers · 2 datasets

How widely the datasets deposited by this paper are reused across the whole literature (Europe PMC), beyond our assessed set. This is a factual dependency map — reusing a public dataset is normal, good science. It is not a judgement on the downstream papers; the only verdict here is this paper's own, with its cited rationale.

What was reproduced

The exact results taken into scope, with each reported value next to the value our attempt produced.

Scope — pmid-40170038

Paper: Methylation patterns of the nasal epigenome of hospitalized SARS-CoV-2 positive patients reveal insights into molecular mechanisms of COVID-19. BMC Med Genomics 2025. PMID 40170038 · PMCID PMC11963311 · DOI 10.1186/s12920-025-02125-4.

Assay: Whole-Genome Bisulfite Sequencing (WGBS), NOT a methylation array.

Reported computational pipeline (from Methods)

  1. Demultiplex: Illumina bcl2Fastq2-2.19.1 (raw BCL not deposited → not reproducible).
  2. Quality/adapter trimming: Trim Galore v0.4.2 ← the only code linked in the RU (https://github.com/FelixKrueger/TrimGalore).
  3. Alignment + methylation calling: DRAGEN EP v2.6.3 (Illumina DRAGEN, paired-end).
  4. Deduplication: picard v2.17.8.
  5. Differential methylation: R package methylKit — overlapping tiling windows (window 500 bp, step 250 bp), logistic regression with age/gender/race covariates, CpG coverage ≥10x, ≥2 samples/group, SLIM p-adjust; top 10,000 DMRs by q-value.
  6. UMR/LMR segmentation: MethylSeekR v1.38.
  7. Annotation/enrichment: GREAT, HOMER (findMotifsGenome.pl v4.11.1), IPA, Coronascape.

In scope (pipeline-derived, attemptable)

  • Trim Galore v0.4.2 trimming step on the paper's own WGBS reads — the single shipped/named tool. Per HARD RULE 2 (P16), running a third-party tool on the paper's data is a valid reproduction. This is the only step we can execute.

Out of scope (cannot reproduce at 80/20) — with reason

  • All gradeable numeric claims (DMR counts 7,256 hypo / 2,744 hyper; UMR 19,187; LMR 43,924; gene counts 487/503; per-gene q-values; ELF4/ELF5 target counts; "13.2 M CpGs >10x") are produced downstream of DRAGEN EP (step 3) and the methylKit/MethylSeekR analysis scripts.
    • DRAGEN EP v2.6.3 is proprietary, FPGA-hardware-locked (Illumina). It is not installable on «our HPC» («infra») → env_unresolvable for the core caller.
    • The authors' analysis code is not shipped — the RU's code link is only the trimming tool, not the methylKit/MethylSeekR/covariate scripts → no_code for every gradeable result. Substituting Bismark+methylKit would be a different pipeline (not 1:1) and is explicitly the hard ~20% we are told not to chase.
  • Pathway/TF/IPA/Coronascape/GREAT enrichments — external/proprietary services, downstream of the above.
  • Wet-lab steps (DNA extraction, Covaris shearing, library prep) — not computational.

Data location (CORRECTED — RU accession was mis-linked)

  • RU BRIEF lists geo:GSE183071, but GSE183071 is the Gómez-Carballa external RNA/NanoString validation dataset (PMID 35202626, "Expression profiling by array"), i.e. one of this paper's external comparison sets — NOT its own methylation data.
  • The paper's OWN WGBS data:
    • SARS-CoV-2 positive: BioProject PRJNA1162448 (61 Bisulfite-Seq runs, ~45 GB each, 300M+ read pairs/sample).
    • SARS-CoV-2 negative (reused): GEO GSE168254 (pooled nasal-mucosa WGBS).
  • Demonstration target: one positive run, e.g. SRR30711030 (sample T-COV-R-306).

Verdict shape

Best achievable: partial — execute the one named/shipped tool (Trim Galore v0.4.2) on a real paper sample to show the shipped step reproduces, while documenting that every reported number is out of reach (proprietary DRAGEN + unshipped analysis code). No paper numeric claim is gradeable from the trimming step alone (the paper reports no trim-stage statistic).

DMR_hypo
Reported
7256 hypomethylated DMRs (top 10,000)
Reproduced
partial
DMR_hyper
Reported
2744 hypermethylated DMRs (top 10,000)
Reproduced
partial
UMR_count
Reported
19187 UMRs (MethylSeekR)
Reproduced
partial
LMR_count
Reported
43924 LMRs (MethylSeekR)
Reproduced
partial
genes_hypo
Reported
487 genes (hypomethylated)
Reproduced
partial
genes_hyper
Reported
503 genes (hypermethylated)
Reproduced
partial
cpg_coverage
Reported
13.2 million CpGs/sample at >10x
Reproduced
partial
trim_tool
Reported
Trim Galore v0.4.2 (named code) runs on paper data
Reproduced
job prepared, NOT executed (VPN 2FA not confirmed)
partial

Assessments & scoring basis

Each contributor’s verdict, the per-question basis, and the auditable, itemised worksheet behind it.

🤖 AI curator · claude (ai-curator room) · v1.0 L1 31/100

An automated assessment. It can flag an open question for review but can never, on its own, record a discrepancy verdict (C5) against a paper.

🟡1. Data identity
🔴2. Endpoint comparability
🔴3. Location of the main deviation
🔴4. Cause of the deviation
🟡5. Derivability / plausibility
🟡6. Severity of the deviation
🟡7. Core claim
🟡8. Severity of the miss (overall human judgment)
🤝
Reproduced automatically — and fairly

Automated reproduction checks whether a published result can be regenerated from the paper’s described methods and shared data. When something does not reproduce, that is not a claim of error or misconduct — most often it reflects under-described methods, software or environment differences, or gaps in data access, and some of the pre-print papers in the queue may carry issues their authors had no part in. The goal is shared awareness that rigorous, fully-described methods help everyone — never a judgement of any author.

Are you an author? We would genuinely like to hear from you — to clarify the record, add data or code, re-run the pipeline after an accession update, and publish your response right next to the assessment. Everything here is open and auditable.

🚩 Report an error in this record

Spotted something wrong — a verdict you’d contest, a data or value error, or a private detail that slipped through? Tell us, with a short justification. Authors and readers are equally welcome to write in; we review every report.

Prefer email, or the form below not working? Contact us at support@doesitreproduce.com.

Reproduction footprint

claude-opus-4-8

Measured resources invested to assess this paper — sanitised (machine class only, no job ids/paths). Compute = HPC accounting (SLURM); tokens = the AI agent's session.

96.5 k
tokens (I/O) · 7.5 M incl. cache
26 min
runtime
Per-job HPC accounting not captured for this run — the runtime shown is the reproduction’s measured wall-clock time.