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

Recurrent RNA edits in human preimplantation potentially enhance maternal mRNA clearance.

Commun Biol · 2022
L1 100/100 3/4
Why this verdict

The main results reproduced: recomputed values matched the published ones within tolerance.

Reproduced on the brainbox compute brainarbeit.com
✓ What held up
  • Reported values were directly comparable
What did not (or only partly)
  • 🟡Could not use the authors’ exact input data
  • 🟡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
100/100
Reproducibility score
1.5 SD above mean
vs. all fields · 1173 studies
🎯 Scores higher than 95% of all assessed papers rank 1 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

DESCRIBED WELL ENOUGH AND REPRODUCED 1:1. This is a meta-analysis (2071 reused RNA-seq samples from 18 GEO datasets; no new wet-lab data). Correct code repo is gao-lab/HERE (the brief's EngeLab/DNTRseq was wrong); primary auditable object is the deposited editome at Zenodo 6658521. Verified 8 pipeline-derived numbers DIRECTLY against the authors' own deposited outputs - all EXACT: C1 989,191 editing sites (distinct sites in the deposited editome), C2 2071 samples = 1797 normal + 274 abnormal across 18 datasets in 29 stage groups, C3 thousands of REEs per early stage (Fig-2 source data reproduced), C4 75.0% of REEs in 3'-UTR / 77.2% exonic, C5 107 edits on 76 genes lost in abnormal embryos. No fabrication flags: every checked value is directly derivable from the shipped data/code. NOT attempted: independent raw-FASTQ re-run of the whole pipeline. Grades are provisional pending human sign-off (see AUDIT.md).

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: 2026-06-19 ⛓ 7a5f3a8e9d47
✎ 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
2026-06-29
Rubric version
v1.0
Assessed by
🤖 AI curator · claude (ai-curator room) · v1.0 · run #1 2026-06-19
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: sonnet
Founding hypothesis

Whether A-to-I RNA editing occurs in a consistent, recurrent pattern across human preimplantation embryos and functionally supports early embryonic development, particularly by enhancing maternal mRNA clearance.

Core claims
  • Compiled the largest human embryonic A-to-I editome to date from 2071 RNA-seq transcriptomes and identified thousands of per-stage Recurrent Embryonic Edits (REEs, present in ≥50% of samples per stage) resource
  • REEs preferentially localize to exonic (mostly 3'-UTR) regions consistently across developmental stages finding
  • REE-targeted genes are recurrently enriched for DNA replication-related functions across ≥3 embryonic stages finding
  • A subset of REE-matching edits (107 edits on 76 genes) undergo organized/complete loss in abnormal (uniparental disomy) embryos and embryos from elder mothers finding
  • REEs likely enhance maternal mRNA clearance finding
  • A proposed mechanism for enhanced clearance is that REEs introduce additional microRNA binding sites (MBSs) into 3'-UTRs of clearance target transcripts mechanism
  • A stringent RNA-seq-only editing identification pipeline (no matched genotypes) was developed and validated, yielding zero overlap between called RNA edits and DNA variants in a paired single-cell DNA-RNA dataset method
  • Editing level of REEs is weakly but significantly negatively correlated with expression level of their target genes across most stage transitions finding
Experimental setups
Assay System Perturbation Readout Platform
RNA-seq (editome identification pipeline) human preimplantation embryos / oocytes, multiple developmental stages (29 groups) none (normal samples) A-to-I RNA editing sites (REEs) genome-wide
Paired DNA-seq and RNA-seq single human cells (independent validation dataset) none overlap ratio of RNA-called edits with matched DNA variants
RNA-seq human uniparental disomy (androgenetic/parthenogenetic) embryos, dataset GSE133854 uniparental disomy (pathological) loss of REE-matching edits
RNA-seq human embryos from elder mothers, dataset GSE95477 advanced maternal age loss of REE-matching edits
RNA-seq human embryos treated with amanitin, dataset GSE101571 amanitin drug treatment abnormal/non-control editing sites
Gene ontology / functional enrichment analysis REE-targeted genes and AG-lost REE-targeted genes (in silico) none enriched biological processes/functions (BH-adjusted p-values)
microRNA binding site (MBS) annotation and statistical testing editing-targeted transcripts (edited inosine treated as guanosine), in silico none MBS gain/loss on 3'-UTR edited transcripts, association with maternal clearance targets
Key results
  • Total of 989,191 editing sites identified across all 2071 samples (normal and abnormal)
  • Validation pipeline yielded zero ratio of identified RNA edits overlapping DNA variants in matched single cells
  • >50% of REEs persisted through stage transitions up to the 2-cell stage; ~30% persisted through the 2-to-4-cell transition >50%; ~30%
  • Cross-stage enriched functions of REE-targeted (3'-UTR) genes were related to DNA replication
  • 107 REE-matching edits on 76 genes were REEs in normal embryos but completely lost in pathological (UPD) embryos and embryos from elder mothers 107 edits/76 genes
  • 3'-UTR REEs were more likely to gain MBSs when overlapping an MBS than general 3'-UTR edits ~50% vs ~33%
  • REEs were more likely to result in MBS gains than in losses of preexisting MBSs (one-tailed paired Wilcoxon test) pseudomedian 1.000017, 95% CI [1.000001, +∞)
  • Genes targeted by maternal mRNA clearance ('decay at 8-cell') had more MBS-gaining REEs than other maternal genes pseudomedian 8.957086×10⁻⁵, 95% CI [8.636118×10⁻⁵, +∞)
Key statistics
  • count 989,191 (total editing sites identified across all curated samples)
  • count 2071 samples in 29 groups (curated human embryonic RNA-seq dataset)
  • correlation between -0.21 and -0.08 (REE editing level vs. target gene expression level across most stage transitions)
  • other ~50% vs ~33% (proportion of MBS-overlapping edits that are MBS-gaining, REEs vs. general 3'-UTR edits)
  • count 107 edits on 76 genes (REEs completely lost in abnormal/elder-mother embryos)
  • other pseudomedian 1.000017, 95% CI [1.000001, +∞) (paired Wilcoxon test, MBS gain vs. MBS loss per (gene, sample) pair)
  • other pseudomedian 8.957086×10⁻⁵, 95% CI [8.636118×10⁻⁵, +∞) (unpaired Wilcoxon test, decay-at-8-cell vs. other maternal genes MBS-gaining REEs (17,060 vs 17,755 pairs))
  • other pseudomedian 1.499948, 95% CI [1.499974, +∞) (unpaired Wilcoxon test, decay-at-8-cell genes vs. 0 baseline)

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.

This is a large-scale computational/bioinformatics study that compiled and reanalyzed 2071 public human embryonic RNA-seq samples to identify recurrent RNA-editing sites and their downstream associations (with gene expression, functional enrichment, and microRNA binding sites). Statistical comparisons were largely nonparametric (Wilcoxon rank-sum tests, chi-square tests) applied to counts or per-(gene,sample) pairs, with results reported as pseudomedians, 95% confidence intervals, and Benjamini-Hochberg (BH)-adjusted p-values for enrichment analyses; some comparisons were explicitly reported with unadjusted p-values.

Replicationunclear Sample size2071 total RNA-seq samples across 29 groups (developmental stage/cell type), comprising 1797 normal and 274 pathological/non-control samples; specific statistical comparisons based on counts of edits/genes or gene-sample pairs (e.g., 36,281 pairs for Fig. 5c; 17,060 vs. 17,755 pairs for Fig. 5d) Groupsnormal vs. pathological/abnormal embryos (e.g., uniparental disomy) and embryos from elder mothers; REEs vs. general edits; MBS gain vs. MBS loss; maternal clearance targets vs. other maternal genes; developmental stages Pairingmixed Randomization/blindingna Dispersionmixed Effect sizesyes Confidence intervalsyes Multiplicity correctionBenjamini-Hochberg (BH) FDR adjustment
Statistical tests used
Test Applied to n Assumptions
Chi-square test (frequency of MBS-gaining edits between MBS-overlapping REEs vs. all MBS-overlapping edits) Fig. 5b counts of site-gained/non-site-gained edits and REEs per stage (Supplementary Data 25); exact n not given in main text not stated
One-tailed, paired Wilcoxon rank-sum test Fig. 5c (MBS gain vs. MBS loss per gene-sample pair) 36,281 gene-sample pairs (Supplementary Data 26) not stated
One-tailed, unpaired Wilcoxon rank-sum test Fig. 5d ("decay at 8-cell" vs. "others", and vs. a 0 baseline) 17,060 pairs (decay at 8-cell) vs. 17,755 pairs (others) not stated
Functional/gene-set enrichment analysis (method not specified) Fig. 3d (REE-targeted gene functions) and Fig. 4b (AG-lost REE-targeted gene functions) gene sets per stage/group; exact n not detailed in main text not stated
Correlation analysis (type not specified) between REE editing level and target gene expression level Supplementary Fig. 22, across stage transitions not stated in main text not stated
Approaches that could also have been used
  • Comparisons of MBS gain vs. loss and of clearance targets vs. other genes were made using one-tailed, nonparametric Wilcoxon rank-sum tests on (gene, sample) pairs.
    Could also: A linear mixed-effects model with gene and/or sample as random effects — Because many pairs likely come from the same genes or samples (non-independent observations), a mixed model could also account for this clustering/hierarchical structure while still testing for a directional shift.
  • Some p-values (e.g., Fig. 5b enrichment analyses) were Benjamini-Hochberg adjusted, while others (Fig. 5d) were explicitly left unadjusted.
    Could also: A single, pre-specified multiplicity-correction scheme (e.g., BH-FDR or Bonferroni) applied uniformly across all reported comparisons — Applying one correction method consistently across the full family of tests is another standard way to control the overall false-positive rate when many comparisons are reported together.
  • The association between REE editing level and target gene expression level was described as a 'statistically significant negative correlation' without specifying the correlation method.
    Could also: Spearman's rank correlation (if a monotonic but non-linear relationship is suspected) or a mixed-effects regression across stages — Spearman correlation is a standard alternative for editing/expression relationships that may not be strictly linear, and a regression framework could also let the association be estimated jointly across all stages rather than stage-by-stage.
  • Group differences (e.g., MBS gain vs. loss, decay-at-8-cell vs. others) were tested with one-tailed hypotheses.
    Could also: Two-tailed Wilcoxon rank-sum tests — A two-tailed test could also be used to remain agnostic to the direction of the effect, which is sometimes preferred when the directionality is a hypothesis being tested rather than an established prior.
  • Functional/gene-set enrichment was performed without stating the specific enrichment test (e.g., hypergeometric, Fisher's exact) in the excerpted text.
    Could also: A named enrichment framework such as a hypergeometric test, Fisher's exact test, or gene set enrichment analysis (GSEA) — Explicitly naming and citing the enrichment method/tool used would let readers evaluate the underlying statistical model (e.g., background gene set definition, one-sided vs. two-sided testing) alongside the BH-adjusted p-values already reported.
  • Frequency of MBS-gaining edits between REEs and general edits was tested with a chi-square test.
    Could also: Logistic regression with edit/REE status as predictor and stage or other covariates included — A regression-based approach could also allow additional covariates (e.g., developmental stage, gene expression level) to be adjusted for simultaneously while estimating the same association.
Software: R · R package ggplot2 (for boxplot conventions)

What was reproduced

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

Scope — pmid-36543858

Paper: Ding Y, Zheng Y, Wang J, et al. "Recurrent RNA edits in human preimplantation potentially enhance maternal mRNA clearance." Commun Biol 2022. PMCID PMC9772385 · DOI 10.1038/s42003-022-04338-0.

Metadata corrections to the harvested brief

  • Code repo in brief was WRONG. Brief listed github.com/EngeLab/DNTRseq (the Enge-lab DNTR-seq method paper — a different study). The paper's actual Code-availability statement points to github.com/gao-lab/HERE (Human Embryonic RNA Editome) + Zenodo 10.5281/zenodo.7386496 (code archive). Reproduction uses the correct repo gao-lab/HERE.
  • Data in brief (GSE133854) is only 1 of 18 reused GEO datasets. The paper compiles 2071 RNA-seq samples from 18 public GEO datasets (all reused, none newly generated wet-lab data). The compiled editome + intermediate results are deposited at Zenodo 10.5281/zenodo.6658521 (the primary auditable data object) and IGV/supp at 10.5281/zenodo.7379397.

Pipeline (from Methods + repo README)

A-to-I (A→G) RNA editing detection, GTEx-style:

  • Trim Galore 0.6.6 → STAR 2.7.0d (splice-aware, hg38 + GENCODE v32) → GATK 3.6.0 variant calling → filter A→G against dbSNP151 / 1000 Genomes / gnomAD / ALFA → SnpEff (GENCODE v32) annotation. Workflow = Snakemake (pipeline.v3*.smk, analysis.v1*.smk). miRNA-binding-site prediction via TargetScan v7.0 ∩ miRanda v1.9.

In scope (pipeline-derived, attemptable)

id result pipeline how we verify
C1 989,191 total A→I editing sites full HERE pipeline count distinct A→G sites in deposited editome (Zenodo editome.files.tar.gz) — authors' own deposited output
C2 2071 samples (1797 normal + 274 abnormal), 18 datasets, 29 stage groups sample compilation count distinct samples / phenotype labels in deposited tables
C3 thousands of REEs (edits in ≥50% of samples per stage) recurrence calc count rows in deposited RE.files/recurrent-edit table
C4 >50% of REEs in 3'-UTR SnpEff annotation tabulate region annotation of REE table
C5 107 REE-matching edits on 76 genes lost in abnormal embryos downstream count in deposited downstream table
C6 (stretch) re-run HERE pipeline on ONE GEO dataset (GSE133854) to call A→G sites and compare site overlap with deposit STAR+GATK slice «our HPC» SLURM job

Primary strategy (honest 1:1): C1–C5 are verified directly against the authors' deposited editome (Zenodo 6658521) — their own pipeline outputs. This is a clear, low-ambiguity data-point reproduction (P16-style: validating the deposited result object). C6 is the harder "keep-going" target: independently re-run the published Snakemake pipeline on one dataset and compare.

Out of scope

  • Full 2071-sample re-run from raw FASTQ (thousands of CPU-hours, ~TBs of SRA download) — not the quick minimum; partially approached via C6 single-dataset slice.
  • Wet-lab A375 DNA/RNA editing validation (experimental) — out of scope.
  • Biological interpretation (maternal mRNA clearance hypothesis) — not a pipeline output.
Figures / tables: Fig 1eFig 1aFig 2aFig 2cFig 4a
C1
Reported
989,191 A-to-I (A>G) editing sites
Reproduced
989191 distinct sites
exact
C2a
Reported
2071 samples
Reproduced
2071
exact
C2b
Reported
1797 normal + 274 abnormal
Reproduced
1797 + 274
exact
C2c
Reported
18 GEO datasets
Reproduced
18
exact
C2d
Reported
29 developmental-stage groups
Reproduced
29
exact
C3
Reported
thousands of REEs per early embryonic stage
Reproduced
oocyte.GV 6383, oocyte.MII 4880, 2-cell 3713, zygote 3078, 4-cell 1384; 36802 distinct REE sites
exact
C4
Reported
REEs mostly 3'-UTR (>75% exonic)
Reproduced
75.0% 3'-UTR (454955/606583); 77.2% exonic
exact
C5
Reported
107 edits on 76 genes lost in abnormal embryos
Reproduced
107 distinct edit positions on 76 genes
exact

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 100/100

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.

74.1 k
tokens (I/O) · 4.3 M incl. cache
12 min
runtime
Per-job HPC accounting not captured for this run — the runtime shown is the reproduction’s measured wall-clock time.