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

CHD7 regulates otic lineage specification and hair cell differentiation in human inner ear organoids.

Nat Commun · 2022
L1 50/100 3/4
Why this verdict

The main results reproduced, with only marginal, non-material deviations.

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
How its reproducibility compares
50/100
Reproducibility score
1.4 SD below mean
vs. all fields · 1173 studies
🎯 Scores higher than 8% of all assessed papers rank 1026 of 1173 scored

A 0–100 reproducibility-quality score from the per-question grades, shown as a z-score: standard deviations above (+) or below (−) the mean of comparable assessments.

Reproduction agent’s raw note

scRNA-seq paper (GSE208585, 12 samples, 10x). Described well enough: YES - Methods name the full pipeline (Seurat v4.0.3 SCTransform/PCA-30/FindClusters/FindMarkers; DESeq2+zingeR DE; EnhancedVolcano) and GEO ships Cell Ranger filtered count matrices, so the analysis boundary is clearly reproducible from deposited data. ACHIEVED on «our HPC»/«infra»: full data download + SHA256 + extraction; exact per-sample called-cell counts. Strong integrity signal - the three merged-dataset raw cell sums BRACKET the paper's reported post-QC totals (25494->22390, 45395->41833, 38274->34703; raw exceeds reported by 8-12%, the expected QC-attrition direction), consistent with no fabrication on the checkable quantities. NOT COMPLETED (graded partial): exact post-QC counts and the central d20 DE result (323 up/129 down) + Fig 4a EnhancedVolcano figure - the Seurat+DESeq2+EnhancedVolcano conda env failed to solve on «our HPC» across 5 attempts (env_unresolvable); the pipeline is fully scripted+staged (reproduction/analysis.R, run.sbatch) and re-runnable once an env is built (Apptainer Bioconductor or scanpy recommended). NOT ATTEMPTED (the hard 20%): hair/supporting-cell type counts (9884/6273), iDEA GSEA, GO enrichment - subjective annotation / non-shipped pipelines. Finalized on operator instruction without further compute. Outcome: PARTIAL - data + integrity checkpoint reproduced 1:1; analysis-derived numeric claims blocked by env build, not by data/code unavailability.

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 Score 50
    assessed: 2026-06-15 ⛓ 1c8f6041e7c1
✎ 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
2026-08-05

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

How do CHD7 mutations (as in CHARGE syndrome) affect human inner ear development, otic lineage specification, and hair cell differentiation, modeled in human pluripotent stem cell-derived inner ear organoids?

Core claims
  • Loss of CHD7 or its chromatin remodeling (ATPase) activity causes complete absence of hair cells and supporting cells in inner ear organoids. finding
  • CHD7 is required for proper otic lineage specification and hair cell differentiation in humans. mechanism
  • Loss of CHD7 dysregulates key otic development-associated genes in mutant otic progenitors. finding
  • CHD7 can regulate certain otic genes (e.g., FBXO2) through a chromatin remodeling-independent mechanism. mechanism
  • Disruption of deafness gene expression in hair cells is a potential mechanism underlying CHARGE-associated sensorineural hearing loss. finding
  • Co-differentiating CHD7 knockout and wild-type cells in chimeric organoids partially rescues mutant phenotypes by restoring dysregulated otic genes. finding
  • Human inner ear organoids from CHD7 mutant hESC lines recapitulate CHARGE syndrome inner ear pathogenesis. method
  • CHD7 is expressed throughout all key otic developmental stages (NNE, OEPD, otic pits, otic vesicles) and in hair cells (high) and supporting cells (low). finding
Experimental setups
Assay System Perturbation Readout Platform
Western blot WT (PAX2nG), CHD7-3xFlag, CHD7 KO/+, CHD7 KO/KO hESCs CRISPR Flag-tagging / CHD7 knockout CHD7 protein expression anti-Flag and anti-CHD7 antibodies
Immunostaining / immunofluorescence Human inner ear organoids (CHD7-3xFlag PAX2nG and CHD7 mutant lines), d20 and d70 CHD7 KO/+, KO/KO, S834F/+, S834F/S834F mutations Marker expression (PAX2, PAX8, EPCAM, MYO7A, SOX2, POU4F3, SOX10, COL9A2, FBXO2, SIX1, DLX5, HOXB9, F-actin stereocilia)
scRNA-seq FACS-isolated PAX2nG+ cells from d20 WT and CHD7 KO/KO organoids CHD7 KO/KO Single-cell transcriptome / differential gene expression in otic progenitors
scRNA-seq FACS-isolated POU4F3nT+/- cells from d70 WT and CHD7 KO/+ organoids CHD7 KO/+ Transcriptome of hair cells and supporting cells
scRNA-seq PAX2nG+ and PAX2nG- populations of d20 WT organoids none Cell-type identity of non-otic populations
Sanger sequencing CHD7 KO/+, KO/KO, S834F mutant hESC alleles (TOPO clones) CRISPR indel / base-editor S834F mutation Allele genotype / mutation confirmation CRISPR base editor
Live imaging d20 and d25 WT and CHD7 KO/KO organoids CHD7 KO/KO Organoid morphology / vesicle expansion
Key results
  • No hair cells or supporting cells observed in CHD7 KO/KO organoids despite normal-looking otic vesicles n=233 aggregates from 7 cultures
  • CHD7 S834F/S834F mutant organoids phenocopied KO/KO, lacking hair cells and supporting cells n=210 aggregates from 5 cultures
  • 323 genes upregulated and 129 downregulated in CHD7 KO/KO otic progenitors vs WT fold change ≥ 2.0, P ≤ 1×10^-10
  • 15 OtoSCOPE deafness genes (incl. TBX1, LMX1A, SOX10) downregulated in CHD7 KO/KO otic progenitors
  • 38.5% of otic lineage-specific genes (incl. FBXO2, COL9A2, OC90) downregulated in KO/KO progenitors, indicating partial loss of otic identity 38.5%
  • Otic progenitors reduced from 89.6% (WT) to 39.7% in CHD7 KO/KO PAX2nG+ cells 89.6% to 39.7%
  • FBXO2 regulated solely by chromatin remodeling-independent mechanism (KO/KO vs S834F/S834F differential)
  • Majority of DE genes in d70 CHD7 KO/+ hair cells downregulated, including deafness genes SIX1, USH1C, STRC fold change ≥ 2.0, P ≤ 1×10^-10
Key statistics
  • count 323 upregulated, 129 downregulated genes (DE genes in CHD7 KO/KO otic progenitors vs WT (fold change ≥ 2.0, P ≤ 1×10^-10))
  • count n = 233 aggregates from 7 independent organoid cultures (CHD7 KO/KO organoids lacking hair/supporting cells)
  • count n = 210 aggregates from 5 independent organoid cultures (CHD7 S834F/S834F organoids lacking hair/supporting cells)
  • count 89.6% (fraction of WT PAX2nG+ cells that are otic progenitors)
  • count 39.7% (fraction of CHD7 KO/KO PAX2nG+ cells that are otic progenitors)
  • count 9,884 hair cells (28.5%) and 6,273 supporting cells (18.1%) (merged d70 WT and CHD7 KO/+ scRNA-seq dataset)
  • count 22,390 cells (d20 WT and CHD7 KO/KO scRNA-seq UMAP dataset)
  • count 14,371 nuclei and cells from 105 otic vesicles (136.87 per vesicle avg) (immunofluorescence intensity quantification across genotypes)

Statistical methods review

Model: opus

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 uses human pluripotent stem cell-derived inner ear organoids with multiple CHD7 mutant lines (KO/+, KO/KO, S834F/+, S834F/S834F) compared to a WT (PAX2nG) control, characterized largely by immunostaining and single-cell RNA-seq. Differential gene expression from scRNA-seq was tested with DESeq2 (two-sided) using fold-change and p-value thresholds, gene set enrichment with the iDEA pipeline (one-sided), and quantified immunofluorescence intensities across genotypes with a Kruskal–Wallis test followed by Dunn's multiple comparisons. Results are reported largely as cell/aggregate/vesicle counts, volcano/bubble/UMAP/violin plots, with significance shown via p-value thresholds and asterisk tiers.

Replicationmixed Sample sizeReported as counts of aggregates, organoid cultures, cells, nuclei, and otic vesicles (e.g., n = 233 aggregates from 7 independent cultures; n = 210 aggregates from 5 cultures; n = 14,371 nuclei/cells from 105 vesicles, 3 vesicles/3 aggregates/3 independent experiments per genotype); no formal power/sample-size calculation described GroupsWT (PAX2nG) vs CHD7 KO/+, KO/KO, S834F/+, S834F/S834F mutant organoids Pairingunpaired Randomization/blindingnot stated DispersionIQR Exact p-valuesno Effect sizesyes Confidence intervalsno Multiplicity correctionBenjamini-Hochberg FDR for DESeq2 DE analysis; Louis method for iDEA p-value adjustment; Dunn's multiple comparisons test following Kruskal–Wallis
Statistical tests used
Test Applied to n Assumptions
DESeq2 (two-sided test) for differential expression scRNA-seq DE genes in d20 CHD7 KO/KO otic progenitors and d70 hair cells/supporting cells (Figs. 4a–b, 6c–d) single-cell counts; 22,390 cells (d20 dataset), 9,884 hair cells and 6,273 supporting cells (d70) not stated
iDEA (Integrative Differential expression and gene set Enrichment Analysis; one-sided test) gene set enrichment of downregulated genes (Figs. 3c, 4d, 6e–h) not stated
Kruskal–Wallis test followed by Dunn's multiple comparisons test immunofluorescence signal intensity per nucleus/cell across genotypes (Fig. 5 violin plots) 14,371 total nuclei and cells from 105 otic vesicles (3 vesicles/3 aggregates/3 independent experiments per genotype) not stated
Approaches that could also have been used
  • Differential expression significance was reported using fixed thresholds (fold change ≥ 2.0 and P ≤ 1×10^-10) with Benjamini-Hochberg FDR.
    Could also: Reporting the BH-adjusted q-values directly alongside an effect-size shrinkage estimate (e.g., DESeq2 lfcShrink) and a stated adjusted-p cutoff would also identify DE genes. — Presenting adjusted q-values and shrunken fold changes conveys the continuous evidence and stabilizes effect sizes for low-count genes, complementing a hard threshold.
  • Per-cell immunofluorescence intensities (many thousands of nuclei/cells from a few vesicles) were compared with Kruskal–Wallis and Dunn's test, treating individual cells as the unit.
    Could also: A hierarchical/mixed-effects model (or summarizing to a per-vesicle or per-experiment mean before testing) could also be used to account for the nested structure (cells within vesicles within experiments). — Modeling the nesting reflects that biological replication is at the organoid/experiment level and would treat clustered measurements accordingly, which some consider when many cells come from few aggregates.
  • Distributions were summarized with median and quartile lines on violin plots.
    Could also: Adding a 95% confidence interval for the median or bootstrap intervals could also describe the central estimate. — An interval around the summary statistic conveys the precision of the estimate in addition to the spread, which is often informative alongside the full distribution.
  • Phenotype absence (no hair/supporting cells) was described qualitatively with counts of aggregates and independent cultures.
    Could also: Reporting proportions with binomial/Wilson confidence intervals, or a Fisher's exact / mixed-effects logistic comparison across genotypes, could also quantify the categorical outcome. — An interval or test on the proportion gives a quantitative measure of certainty for the all-or-none phenotype across replicates.
  • Gene set enrichment used the iDEA pipeline (one-sided) with Louis-method p-value adjustment.
    Could also: Complementary GSEA approaches (e.g., fgsea/GSEA preranked or a hypergeometric/over-representation test with BH correction) could also be applied. — Cross-checking enrichment with an independent method and a widely-used multiplicity correction can corroborate which pathways are highlighted.
Software: DESeq2 · iDEA (Integrative Differential expression and gene set Enrichment Analysis)

Result convergence & founder nodes

Findings this paper shares with others that ran a comparable experiment. A node’s strength is how many independent papers report it (replication breadth) — not how often it is cited, so a heavily-replicated but under-cited founder still stands out.

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
44
Impact: medium
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.

What was reproduced

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

Scope — pmid-36396635

Paper: Nie J, Ueda Y, Solivais AJ, Hashino E. CHD7 regulates otic lineage specification and hair cell differentiation in human inner ear organoids. Nat Commun 2022;13:7053. PMCID PMC9672366. DOI 10.1038/s41467-022-34759-8.

Data: GEO GSE208585 — scRNA-seq (10x Chromium 3' v3, NovaSeq 6000), 12 samples. GEO ships Cell Ranger filtered count matrices (barcodes/features/matrix.mtx per sample) → analysis can start from the cell×gene matrices (no FASTQ alignment needed).

Code named by paper: Seurat v4.0.3 (Read10X, CreateSeuratObject, QC filter, SCTransform, PCA[30 PCs], FindNeighbors/FindClusters, FindMarkers); DE via DESeq2 + zingeR; visualization via EnhancedVolcano (the repo link in the brief).

Sample → dataset mapping (from GEO titles)

dataset fig GSMs groups reported N cells
d20 PAX2nG+ otic Fig 3b / 4a 6354844 (KO/KO), 6354845 (WT) WT, KO/KO 22,390
d70 EPCAM Fig 5a 6354850–853 WT/KO EPCAM± 41,833 (Cluster0=2810)
d70 POU4F3nT Fig 6a 6354846–849 WT/KO+ nT± 34,703 (HC 9884 / SC 6273)
d20 WT PAX2nG± Fig 1 6354854,855 WT only (not targeted)

IN SCOPE (pipeline-derived, reproducible from shipped matrices)

  • C1 Merged cell counts after QC for the 3 datasets (22,390 / 41,833 / 34,703).
  • C2 d20 otic-progenitor DE: 323 up / 129 down genes (FC≥2, P≤1e-10), Fig 4a — the EnhancedVolcano figure. Primary target.
  • C3 d70 POU4F3nT hair-cell / supporting-cell counts (9884 / 6273).

OUT OF SCOPE (not attempted, with reason)

  • Cell Ranger alignment from FASTQ — FASTQ not on GEO (only matrices shipped); upstream of the reproducible boundary. We start from the shipped filtered matrices.
  • iDEA GSEA bubble plots (Figs 4d, 6g-h), GO enrichment — separate non-shipped pipeline, manual gene-set curation; the hard 20%.
  • Wet-lab claims (Western blot, immunostaining, FACS gating, stereocilia morphology).
  • Exact QC thresholds: paper states only "extremely high/low UMIs, high % mito" with no numeric cutoffs → cell counts are threshold-sensitive (graded with tolerance).
  • zingeR is archived (removed from Bioconductor); if it won't install we substitute Seurat's Wilcoxon DE and flag the method deviation explicitly.

Method-deviation notes

  • Single 10x run per genotype ⇒ no biological replicates ⇒ pseudobulk DESeq2 is n=1 per group (invalid). Cell-level DESeq2+zingeR (cells as replicates) is what the paper used; we reproduce DE with Seurat FindMarkers (Wilcoxon) as the clean, deterministic alternative and report the count of genes passing the paper's FC≥2 & P≤1e-10 gate.
Figures / tables: Fig 3bFig 5aFig 6aFig 4a
C1a
Reported
22,390 d20 PAX2nG+ cells (Fig 3b)
Reproduced
raw deposited 25,494 (12138+13356); post-QC not computed
partial
C1b
Reported
41,833 d70 EPCAM cells (Fig 5a)
Reproduced
raw deposited 45,395 (GSM850-853); post-QC not computed
partial
C1c
Reported
34,703 d70 POU4F3nT cells (Fig 6a)
Reproduced
raw deposited 38,274 (GSM846-849); post-QC not computed
partial
C2up
Reported
323 upregulated genes, d20 otic prog KO vs WT (FC>=2,P<=1e-10; Fig 4a)
Reproduced
not completed - analysis env failed to build
partial
C2dn
Reported
129 downregulated genes (Fig 4a)
Reproduced
not completed - analysis env failed to build
partial
C2fig
Reported
Fig 4a volcano via EnhancedVolcano
Reproduced
not completed - analysis env failed to build
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 50/100

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.

171.8 k
tokens (I/O) · 10.1 M incl. cache
27 min
runtime · 0.09 CPU-h
4.3 GB
peak RAM
3 (3 failed)
HPC jobs
hummel
machine