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

An open RNA-Seq data analysis pipeline tutorial with an example of reprocessing data from a recent Zika virus study.

F1000Res · 2016
L1 74/100 PQI 92
Why this verdict

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

Reproduced on the brainbox compute brainarbeit.com
Scoring basis — itemised

Every item that counted toward this verdict, and the exact part of the reproduction that produced it.

Supporting (toward a concern)
Content-critical question only partially held
+2 pts
From: Q5 · Derivability / plausibility 🟡
Content-critical question only partially held
+2 pts
From: Q7 · Core claim 🟡
Content-critical question only partially held
+2 pts
From: Q8 · Severity of the miss (overall human judgment) 🟡
Minor / cosmetic deviation
+1 pts
From: Q3 · Location of the main deviation 🟡
Concordant (toward reproduced)
Code + data deposited & functional
-2 pts
From: Data & code availability Available & functional
Total score +5
✓ What held up
  • Same input data as the authors
  • Reported values were directly comparable
  • No authors-side cause for any deviation
  • Any deviation was negligible
What did not (or only partly)
  • 🟡A deviation arose in the data or preprocessing
  • 🟡Reported values were not (fully) derivable from the shared data
  • 🟡The central claim did not (fully) hold under reproduction
  • 🟡Overall, the reproduction showed a material discrepancy
How its reproducibility compares
74/100
Reproducibility score
at the mean
vs. all fields · 1173 studies
🎯 Scores higher than 43% of all assessed papers rank 644 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

REPRODUCED. Described well enough to reproduce; authors' own tutorial repo (P16) re-run end-to-end on «our HPC» against GSE78711/SRP070895. Full pipeline SRA/ENA->STAR 2.4.2a (iGenomes UCSC hg19)->featureCounts 1.4.6-p5->edgeR CPM->Characteristic Direction (geode, gamma=0.5)->cross-platform cosine. The HEADLINE claim (two sequencers give near-identical CD differential signatures) reproduces to <1%: cosine distances 0.012345/0.004577/0.003075 vs reported 0.012347/0.004558/0.003052 (C6 essentially exact, -0.014%). Per-sample featureCounts (C4) reproduces to +0.31% (10893451 vs 10859768) and NextSeq spot counts (C5) are exact. Filter shapes C2/C3 within ~1-3%. The ONLY mismatch is the raw matrix gene count C1 (26364 vs 25370, +3.9%), fully explained by the iGenomes UCSC hg19 genes.gtf annotation being revised since 2016 (count is annotation-determined) -- reference drift, NOT a pipeline error or fabrication; C2/C3 shifts track it. NO fabrication suspected: every checked number is derivable from the shipped data+code. Data QC: all 12 fastq.gz md5-verified against ENA (one truncated mate from an aborted download was caught via an anomalous 1.26M Assigned count, re-downloaded, and re-aligned to the correct 10.89M). NOT ATTEMPTED (out of scope, 80/20): Enrichr/L1000CDS2/Clustergrammer external web-API enrichment + drug-repurposing figures (qualitative, third-party services; CD gene directions are biologically consistent). All grades provisional for human audit (AUDIT.md).

💻 Code ↗ 🗄 Data: GSE78711

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 59
    assessed: 2026-06-16 ⛓ 55fc11191d75
✎ 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
2026-06-24
Rubric version
v1.0
Assessed by
🤖 AI curator · claude (ai-curator room) · v1.0 · run #1 2026-06-16
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

Can an open, reproducible RNA-seq analysis pipeline (delivered as an IPython notebook and Docker image) be used to reprocess published RNA-seq data and extract additional biological knowledge, tested here by reanalyzing a Zika virus (ZIKV) infection dataset to probe molecular mechanisms linked to ZIKV-associated microcephaly?

Core claims
  • An open-source, reproducible RNA-seq pipeline delivered as an IPython notebook and Docker image can process raw RNA-seq data into interactive PCA/HC plots, enrichment results, and small-molecule predictions with minimal setup overhead method
  • Downregulated genes after ZIKV infection of neuronal progenitors are enriched for cell cycle-related processes, confirming the original study's finding finding
  • Upregulated genes after ZIKV infection significantly overlap with genes that, when knocked out in mice, cause abnormal brain morphology and abnormal nervous system phenotypes, suggesting a link to microcephaly finding
  • Downregulated genes after ZIKV infection are enriched for targets of the transcription factors E2F4 and FOXM1 finding
  • Querying ZIKV-induced gene expression signatures against LINCS L1000 data via L1000CDS2 predicts small molecules that could mimic or reverse the ZIKV-induced expression changes resource
  • Packaging the pipeline as a Docker image ensures a consistent, static software environment that facilitates reproducibility across computers, clusters, and cloud method
Experimental setups
Assay System Perturbation Readout Platform
bulk RNA-seq human iPSC-derived neuronal progenitor cells ZIKV infection vs mock gene expression levels (CPM, RPKM) MiSeq and NextSeq (Illumina)
principal component analysis (PCA) same RNA-seq samples (8 total: 4 mock, 4 ZIKV-infected) ZIKV infection vs mock sample clustering in first two principal components Python/IPython notebook
hierarchical clustering top 800 highest-variance genes across the 8 samples ZIKV infection vs mock log-transformed, z-score normalized expression clusters Clustergrammer
differential expression analysis (Characteristic Direction method) same RNA-seq dataset ZIKV infection vs mock ranked lists of up/downregulated genes
gene set enrichment analysis differentially expressed gene lists from ZIKV dataset none enriched terms across >180,000 gene sets in 90 libraries (e.g., ChEA, KEGG, MGI Mammalian Phenotype) Enrichr
small molecule signature search LINCS L1000 data (30,000 signatures) none small molecules predicted to mimic or reverse ZIKV expression signature L1000CDS2
Key results
  • Downregulated genes after ZIKV infection are enriched for cell cycle-related genes, consistent with the original study
  • Downregulated genes are enriched for targets of transcription factors E2F4 and FOXM1
  • Top two enriched MGI Mammalian Phenotype Level 4 terms for upregulated genes are 'abnormal nervous system' (MP0003861) and 'abnormal brain morphology' (MP0002152)
  • PCA shows samples cluster both by infection status (ZIKV vs mock) and by sequencing platform (MiSeq vs NextSeq)
  • L1000CDS2 query of ZIKV-induced signatures yields candidate small molecules predicted to mimic or reverse the expression changes
Key statistics
  • count 8 samples (4 untreated, 4 infected) (RNA-seq dataset GSE78711 from ZIKV-infected vs mock neuronal progenitors)
  • count 800 genes (genes with largest variance used for hierarchical clustering heatmap)
  • count over 180,000 annotated gene sets across 90 libraries (Enrichr enrichment analysis database size)
  • count 30,000 significant signatures (LINCS L1000 signatures processed with Characteristic Direction method used by L1000CDS2)
  • other MP0003861 (abnormal nervous system), MP0002152 (abnormal brain morphology) (top enriched MGI Mammalian Phenotype terms for upregulated genes)

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.

This is a methods/pipeline tutorial paper that reprocesses a published RNA-seq dataset (8 samples: 4 ZIKV-infected vs 4 mock, sequenced in duplicate on two platforms). The analytical workflow is descriptive and exploratory rather than inferential: reads were aligned with STAR, quantified with featureCounts, normalized to CPM/RPKM with edgeR, then explored with PCA and hierarchical clustering, with differential expression ranked by the authors' Characteristic Direction (CD) multivariate method and followed by enrichment analysis (Enrichr) and signature search (L1000CDS2). Results are reported as ranked gene lists, interactive PCA/clustering visualizations, and enrichment terms rather than as per-gene test statistics with reported p-values or confidence intervals in the main text.

Replicationmixed Sample sizeStated descriptively as eight samples total (four untreated, four infected), measured in duplicate on two platforms (MiSeq and NextSeq); no formal power/sample-size calculation described GroupsZIKV-infected vs. mock-treated human neuronal progenitors (two conditions) Pairingunclear Randomization/blindingna Dispersionnone Exact p-valuesno Effect sizesno Confidence intervalsno Multiplicity correctionnone stated
Statistical tests used
Test Applied to n Assumptions
Characteristic Direction (CD), a multivariate geometrical method for ranking differentially expressed genes between two conditions identification of differentially expressed genes, infected vs. control 8 samples (4 infected, 4 control), in duplicate across two platforms not stated
Principal component analysis (PCA), exploratory/unsupervised Figure 1, sample-level structure 8 samples na
Hierarchical clustering of the 800 highest-variance genes (log-transformed, z-score normalized) Figure 2 heatmap 8 samples, 800 genes na
Gene set enrichment analysis via Enrichr (over 90 gene set libraries) Figure 3, Table S1, enrichment of up/down gene lists ranked up/down gene lists from CD not stated
Approaches that could also have been used
  • Differential expression was ranked using the Characteristic Direction (CD) method, which produces ranked gene lists.
    Could also: Established count-based DE frameworks such as edgeR (already used here for normalization), DESeq2, or limma-voom, which fit negative-binomial or linear models and report per-gene log-fold-changes with p-values and FDR-adjusted q-values. — These would additionally attach calibrated significance and adjusted p-values to each gene, complementing CD's ranking with an explicit statistical cutoff; CD offers a multivariate, threshold-free ranking that some prefer for downstream signature search.
  • Enrichment terms are described qualitatively as the 'top' or 'most significant', without an explicit multiplicity correction stated in the text.
    Could also: Reporting Benjamini-Hochberg FDR-adjusted q-values for the enrichment results, as Enrichr can output. — Stating the adjusted significance for the family of gene-set tests would make the basis for 'most significant' explicit; one referee raised this same point about whether FDR was applied.
  • The analysis is exploratory/descriptive with n=4 per condition and emphasizes ranked lists and visualizations rather than reported numeric statistics.
    Could also: Summarizing key quantities with measures of spread (SD, IQR, or 95% confidence intervals) and exact p-values where comparisons are made. — Such reporting conveys the magnitude and uncertainty of effects directly in the text and is often preferred for small-n studies; the chosen ranking-plus-visualization approach is well suited to a tutorial/pipeline focus.
  • Samples were collected across two sequencing platforms (MiSeq and NextSeq), and PCA shows clustering by both condition and platform.
    Could also: Explicitly modeling platform as a batch covariate (e.g., in a limma/edgeR design matrix) or applying a batch-correction step such as ComBat/RUVSeq before downstream analysis. — Incorporating the platform effect into the model would separate condition-driven from platform-driven variance in the DE estimates; the authors instead surface the platform structure transparently through PCA and clustering.
  • Hierarchical clustering used the 800 genes with the largest variance, z-score normalized.
    Could also: Sensitivity analyses across different numbers of selected genes, alternative distance metrics/linkage, or consensus clustering. — This would show how stable the sample grouping is to the feature-selection threshold; fixing a single gene count keeps the tutorial simple and reproducible.
Software: STAR (aligner) · featureCounts · edgeR (Bioconductor) · FastQC · Enrichr · L1000CDS2 · Python / IPython (Jupyter) notebook · clustergrammer (web-based visualization)

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
36
Impact: medium
Foundation confidence
None of its references are in our reproducibility record yet — its foundation cannot be assessed.
Topics

Assessed papers, coloured by verdict. Click a node to open it.

Built on (assessed references) (0)
  • No assessed neighbours yet — the network grows as more papers are assessed.
Cited by (assessed papers) (1)

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.

GSE78711 GEO in Methods (http://purl.org/orb/Methods)
no other assessed paper uses this yet
SRP070895 ENA in Methods (http://purl.org/orb/Methods)
no other assessed paper uses this yet

What was reproduced

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

Scope — pmid-27583132

Paper: Wang Z, Ma'ayan A. An open RNA-Seq data analysis pipeline tutorial with an example of reprocessing data from a recent Zika virus study. F1000Research 2016. PMID 27583132 · PMC4972086 · DOI 10.12688/f1000research.9110.1 Code: https://github.com/MaayanLab/Zika-RNAseq-Pipeline (commit fb16b32) Data: GEO GSE78711 / SRA SRP070895 (the Zika study reprocessed) — 8 RNA-seq runs of human neural-progenitor-like cells, mock vs ZIKV-infected, on two platforms:

  • MiSeq, PAIRED: SRR3191542, SRR3191543 (control), SRR3191544, SRR3191545 (Zika)
  • NextSeq 500, SINGLE: SRR3194428, SRR3194429 (control), SRR3194430, SRR3194431 (Zika)

The pipeline (what the tutorial actually does)

.srafastq-dump (sra-tools 2.6.2) → STAR 2.4.1c align to Illumina iGenomes UCSC hg19 (genome.fa + Annotation/Genes/genes.gtf) → featureCounts (subread 1.4.6-p2, -t exon -g gene_id) → edgeR CPM matrix (normalize.R) → Python downstream (RNAseq.py, Zika.ipynb): filter, PCA, Characteristic Direction differential expression (geode.chdir, gamma=0.5), Enrichr / L1000CDS2.

In scope (pipeline-derived; we attempt) — graded by how annotation-vs-data driven

id reported (notebook/tutorial) depends on risk
C1_genecount featureCounts matrix = 25370 genes × 8 GTF only low — fixed by iGenomes hg19 genes.gtf
C2_filt_expr after dropping non-expressed: (21983, 8) counts→CPM med
C3_filt_low after low-expr filter (>0.3 CPM in >2 samples): (16242, 8) counts→CPM med
C4_fc_assigned SRR3191542 featureCounts Assigned=10,859,768 (NoFeatures 3,658,745; MultiMap 1,315,262; Ambig 277,433) STAR+subread+GTF med (version-sensitive)
C5_spots SRA spots: SRR3194428=72,983,243; 429=94,729,809; 430=76,299,868 SRA metadata low — exact
C6_cd_cos1 cosine(NextSeq500, MiSeq) CD sigs = 0.012346509286 full pipeline med (robust metric)
C7_cd_cos2 cosine(NextSeq500, combined) = 0.00455841805086 full pipeline med
C8_cd_cos3 cosine(MiSeq, combined) = 0.00305230907374 full pipeline med
C9_enrich_dir down-genes → E2F4 targets / cell-cycle (KEGG); up-genes → MGI brain-morphology phenotypes CD genes + Enrichr qualitative

The "headline" reproducible computational claim is the cross-platform CD-signature concordance (C6–C8): two different sequencers give near-identical differential signatures (cosine distance ~0.003–0.012, i.e. cosine similarity ~0.99).

Out of scope (the hard/external ~20% — not attempted, by 80/20 rule)

  • Enrichr / L1000CDS2 / Clustergrammer / Harmonizome web-API results and the drug mimicker/reverser figures — external third-party web services, shown only qualitatively as screenshots in the tutorial; not deterministic numbers.
  • 3D plotly PCA, interactive clustergram (visual, no numeric claim).
  • We do NOT re-derive the original Zika study's own results (Tang et al.); we reproduce the tutorial's reprocessing pipeline outputs only.

P16 note

This repo IS the authors' own tutorial code, applied to a third party's data (GSE78711). Reproduction = run the shipped pipeline on the shipped SRA accessions with the shipped reference (iGenomes hg19) and compare the tutorial's printed example numbers. All heavy compute on «our HPC» («infra»), per HARD RULES.

C5_spots
Reported
SRR3194428=72983243; 429=94729809; 430=76299868
Reproduced
identical
exact
C6_cd_cos_NextSeq_MiSeq
Reported
0.012346509286
Reproduced
0.012344771775
within tolerance
C7_cd_cos_NextSeq_combined
Reported
0.00455841805086
Reproduced
0.004577404955
within tolerance
C8_cd_cos_MiSeq_combined
Reported
0.00305230907374
Reproduced
0.003074758120
within tolerance
C4_fc_assigned
Reported
SRR3191542 Assigned=10859768
Reproduced
10893451 (+0.31%)
within tolerance
C3_filt_low
Reported
(16242,8)
Reproduced
(16403,8) (+0.99%)
within tolerance
C2_filt_nonzero
Reported
(21983,8)
Reproduced
(22658,8) (+3.07%)
within tolerance
C1_genecount
Reported
(25370,8)
Reproduced
(26364,8) (+3.92%, iGenomes hg19 annotation drift since 2016)
did not match
C9_enrich_dir
Reported
down: cell-cycle; up: brain-morphology
Reproduced
qualitatively consistent CD gene directions (down FOS/EGR1/proliferation; up ASNS/HSPA5/ER-stress); Enrichr web API not called
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 74/100

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)
Scoring basis — itemised

Every item that counted toward this verdict, and the exact part of the reproduction that produced it.

Supporting (toward a concern)
Content-critical question only partially held
+2 pts
From: Q5 · Derivability / plausibility 🟡
Content-critical question only partially held
+2 pts
From: Q7 · Core claim 🟡
Content-critical question only partially held
+2 pts
From: Q8 · Severity of the miss (overall human judgment) 🟡
Minor / cosmetic deviation
+1 pts
From: Q3 · Location of the main deviation 🟡
Concordant (toward reproduced)
Code + data deposited & functional
-2 pts
From: Data & code availability Available & functional
Total score +5

On the graded numbers this is a strong partial reproduction of the authors' own tutorial pipeline: C5 SRA spot counts are identical and C4 featureCounts assignment reproduces to within 0.31%, cross-validating both alignment and the original printed values, with no fabrication suspected. The single mismatch (C1, 26,364 vs 25,370 genes, +3.9%) sits on the input/annotation side and is fully explained by post-2016 iGenomes hg19 GTF drift — a technical/expected cause, not an authors' defect. The weakness is on our side: the run was finalized early, so the headline cross-platform CD cosine distances (C6–C8) and CPM filter shapes (C2/C3) were never computed, leaving the central concordance claim unconfirmed though plausible. Net: solid with explainable deviations, but incomplete on the headline — yellow overall.

🤝
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.

330.2 k
tokens (I/O) · 26.6 M incl. cache
121 min
runtime · 4.74 CPU-h
34.7 GB
peak RAM
3
HPC jobs
hummel
machine