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

All of gene expression (AOE): An integrated index for public gene expression databases.

PLoS One · 2020
not yet assessed 3/4
⚑ Flagged for review — a reproduced result did not match the reported value

Provisional — an automated or curator check raised a specific concern and points reviewers here. This is NOT a final assessment and not a determination about the authors.

Why this verdict

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

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 🟡
Minor / cosmetic deviation
+1 pts
From: Q6 · Severity of the deviation 🟡
Input / endpoint not comparable 1:1
+1 pts
From: Q1 · Data identity 🔴
Concordant (toward reproduced)
Code + data deposited & functional
-2 pts
From: Data & code availability Available & functional
Total score +7
✓ 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 (no_expected_result). AOE is a database/index RESOURCE paper by a single author (Bono), not a fixed-output computational pipeline. Its repo is real and runnable (Perl5/shell metadata harvesters over ArrayExpress + GEO/SRA building a tab-delimited index, plus a Flask/D3.js web app), but NONE of its reported numbers is a pinnable, reproducible pipeline output: the GEO-vs-ArrayExpress comparison counts (>30k / >10k) are time-varying live-DB snapshots from ~2019, the search-example counts (524 / 58) reflect current live-site DB state, and the usage figures (95,334 visits / 393,174 page views / 630,837 hits) are the author's private awstats server logs (not derivable from any shipped artifact -- flagged per HARD RULE 5, but consistent with normal private analytics, not evidence of fabrication). The only quantitative cross-DB comparison additionally depends on the legacy ArrayExpress REST/MAGE-TAB API, which has been retired and migrated to EBI BioStudies (verified: .../arrayexpress/json/v3/... now 302-redirects to biostudies/arrayexpress help), so its upstream input no longer exists in the original form. The text-mined data accession GSE52334 is a false positive -- it appears in the paper only as a nomenclature example (GSE52334 <-> E-GEOD-52334), not as analysis data, so there is no dataset to run a tool over. Per the 80/20 + honest-1:1 rule, running the pipeline today would yield only a fresh 2026 snapshot, which is NOT a reproduction of the 2019 figures; producing and grading such a number would be misleading, so no result was attempted and no «our HPC» compute was spent. NOT attempted: re-harvesting current GEO/SRA/BioStudies to regenerate index counts (would not be 1:1), and reconstructing private web-usage logs (impossible). Decision reached entirely at control-plane screening with API status checks (evidence in scope.md / code.json / data.json / AUDIT.md).

💻 Code ↗ 🗄 Data: GSE52334

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 ⛓ 9898aafc3223
✎ 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

Public gene expression databases (GEO, ArrayExpress, GEA) are separately maintained and may lack RNA-seq data available in the SRA; an integrated index is needed to make all public gene expression data searchable together.

Core claims
  • AOE integrates publicly available gene expression data from GEO, ArrayExpress, and GEA into a single searchable index. resource
  • AOE includes RNA-seq data from the SRA that are not present in GEO or ArrayExpress, captured via a three-level indexing scheme (levels 1-3). resource
  • ArrayExpress discontinued GEO data import in 2017, leaving over thirty thousand GEO entries missing from ArrayExpress, necessitating direct indexing of GEO. finding
  • AOE provides an interactive graphical web interface (using D3.js histograms) for querying gene expression data by year, organism, and quantification method. method
  • AOE provides a REST API returning JSON output for automated programmatic searches. method
  • The existence of AOE level 3 data shows that not all sequencing gene expression data are stored in GEO, indicating GEO is insufficient as a complete public gene expression database. finding
  • AOE indexes are built by parsing MAGE-TAB (ADF/IDF/SDRF) files from ArrayExpress/GEA and SRA/BioProject/BioSample metadata via the DBCLS SRA API. method
  • AOE serves as the search interface for the DDBJ Genomic Expression Archive (GEA) and is linked from the GEA website. resource
Experimental setups
Assay System Perturbation Readout Platform
Metadata indexing from ArrayExpress (MAGE-TAB: ADF/IDF/SDRF files) EBI ArrayExpress public database none gene expression data-series entries (AOE level 1) AE FTP site
Metadata indexing of GEO data via DBCLS SRA API NCBI GEO / SRA / BioProject / BioSample none merged gene expression data-series entries with BioProject and GEO IDs (AOE level 2) DBCLS SRA API (JSON-LD)
Metadata indexing of RNA-seq data not in GEO via DBCLS SRA API Sequence Read Archive (SRA) none transcriptome sequencing data-series entries with BioProject IDs only (AOE level 3) DBCLS SRA API
Cross-database overlap comparison by matching series IDs ArrayExpress vs GEO entries (GSE vs E-GEOD IDs) none number of overlapping/missing data-series entries
Data visualization / interactive querying AOE web interface keyword/organism/technology filtering histograms and retrievable record counts Python3, D3.js
Key results
  • Over thirty thousand GEO entries were missing in ArrayExpress after import was discontinued over 30,000 entries
  • Over ten thousand entries present in ArrayExpress were missing in GEO over 10,000 entries
  • Keyword search for 'hypoxia' returns 524 items across all data 524 items
  • Filtering 'hypoxia' results to Homo sapiens + Illumina yields 58 records 58 records
  • AOE web server usage over ~4 years (July 2015-Oct 2019) 95,334 visits; 393,174 page views; 630,837 hits
  • Two-thirds of visits were under 30 seconds, indicating instant keyword queries ~2/3 of visits <30 s
Key statistics
  • count over thirty thousand (GEO entries missing in ArrayExpress)
  • count over ten thousand (ArrayExpress entries missing in GEO)
  • count 524 (items returned for 'hypoxia' keyword search)
  • count 58 (records after filtering hypoxia to Homo sapiens and Illumina)
  • count 95,334 (AOE web server visits, July 2015-Oct 2019)
  • count 393,174 (AOE web server page views, July 2015-Oct 2019)
  • count 630,837 (AOE web server hits, July 2015-Oct 2019)
  • count 30 (top species listed via species icon shortcut)

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 database/resource paper describing the construction and utility of All Of gene Expression (AOE), an integrated index for public gene expression databases. The work is descriptive and infrastructural: it reports counts of overlapping data-series entries between ArrayExpress and GEO, web-usage statistics (visits, page views, hits over a time window), and item counts returned by example keyword searches. No inferential statistical hypothesis testing, experimental group comparison, or modeling is reported.

Replicationna GroupsDescriptive comparison of data-series entry overlap between ArrayExpress and GEO databases (Fig 1); no experimental groups Pairingna Randomization/blindingna Dispersionnone
Approaches that could also have been used
  • Database entry overlap between ArrayExpress and GEO is presented as raw counts and a figure (Fig 1).
    Could also: The same overlap could also be summarized with proportions or a Venn/UpSet diagram alongside the counts, and the fraction unique to each source reported with a 95% confidence interval treating it as a proportion. — Reporting proportions with intervals would add a sense of relative magnitude and sampling uncertainty to the descriptive counts, complementing the absolute numbers.
  • Web-server usage is reported as aggregate totals over July 2015–Oct 2019 (visits, page views, hits) plus the statement that two-thirds of visits were under 30 seconds.
    Could also: Usage could also be summarized as a time series (e.g., monthly visits) with a measure of central tendency and spread, and the session-duration distribution shown as a histogram or with median and IQR. — A temporal and distributional view would convey trends and variability in usage over the multi-year window in addition to the single cumulative totals.
  • Example query results are reported as point counts (e.g., 524 items, then 58 records after filtering).
    Could also: These could also be accompanied by the date/version of the underlying index snapshot and the total indexed corpus size, so counts are expressed as a fraction of the whole. — Anchoring counts to a corpus size and snapshot date would make the numbers reproducible and comparable across database updates.
Software: Python3 (visualization scripts) · Perl5 (data parsers) · UNIX shell · D3.js (JavaScript visualization library)

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.

Authors · 1
Citations
38
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.

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.

GSE52334 GEO in Results (http://purl.org/orb/Results)
no other assessed paper uses this yet

Downstream reach in the literature

0 downstream papers · 1 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.

This paper is currently under reproducibility review (see the verdict above). The map below shows where the data in question has propagated — so reuse can be traced, not so the downstream work is presumed affected.
GSE52334 GEO reused by 1 papers in the literature

What was reproduced

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

Scope — pmid-31978081 (AOE: All Of gene Expression)

  • Title: All of gene expression (AOE): An integrated index for public gene expression databases.
  • Author: Bono H. (single author) · PLoS ONE 2020 · DOI 10.1371/journal.pone.0227076
  • Code: https://github.com/dbcls/AOE (pinned commit 140ba7fcc81e4370ac007bd94e4441213949cbff, pushed 2022-04-15)
  • "Data" accession harvested: GEO GSE52334 — see note below (text-mining false positive).

What the paper is

AOE is a resource / index paper. It describes building an index over public transcriptome databases by harvesting metadata from ArrayExpress (DDBJ MAGE-TAB mirror), NCBI GEO (via the DBCLS SRA API), and SRA (SRA_Accessions.tab), merging them into one tab-delimited table, and serving a faceted web front-end (Flask + D3.js / crossfilter). Methods (verbatim): parsers "implemented in Perl5 and UNIX shell commands"; visualization "Python3 scripts … D3.js"; usage analytics via awstats. No statistical analysis, no differential expression, no dimensionality reduction, no machine-learning pipeline is described.

In-scope vs out-of-scope (per HARD RULE 2)

Reported result (location) Pipeline that would produce it In scope? Why
">thirty thousand entries missing in AE", ">ten thousand AE entries missing in GEO" (Results, GEO↔AE comparison) 02GEO-AE.pl + AE mirror + GEO/SRA API NO Live-DB snapshot at ~2019. No shipped expected value; not pinnable. Upstream ArrayExpress legacy API retired (now BioStudies) → input source gone. A 2026 snapshot vs a 2019 snapshot of a deliberately moving target is not a 1:1 reproduction.
Search example "524 items … 58 records (Homo sapiens + Illumina)" (Results / Fig) live AOE web query NO Reflects current DB state of the live site; not a fixed computed value.
Usage stats: "95,334 visits, 393,174 page views, 630,837 hits" (Jul 2015–Oct 2019) author's private awstats server logs NO Not derivable from any shipped data or code (private web-server logs). Flagged per HARD RULE 5 — irreproducible by anyone but the author. (No evidence of fabrication; this is normal private analytics, simply non-auditable.)
Nomenclature: "GSE52334 in GEO corresponds to E-GEOD-52334 in AE" naming convention (s/E-GEOD-(\d+)/GSE$1/) trivial True by construction; a naming rule, not a computational result. This is the only mention of GSE52334 — it is not an analysis dataset.
Index schema / column layout (README table, API_documentation.md) 01AE2REI.pl, 03mkAOEtab.pl descriptive Format spec, no reported numeric value to compare.

Net: zero in-scope, pinnable, reproducible pipeline outputs.

Why this is a drop (not a partial)

  1. No expected result that is pinnable — every reported figure is a time-varying snapshot of live external databases or a private web-server log. There is no shipped reference output, no fixed expected value, and the paper itself reports approximate counts ("over thirty thousand") of a moving target.
  2. Upstream data source retired — the ArrayExpress side of the only quantitative comparison depends on the legacy ArrayExpress REST/MAGE-TAB service, which has been migrated to EBI BioStudies; the v3 JSON endpoints the scripts parse now redirect to biostudies/arrayexpress help. The DDBJ AE mirror + author-internal hosts (gw.ddbj.nig.ac.jp, dtn4, «path», aoe2018 EC2) referenced by 99updateAOEall.sh are unavailable.
  3. Harvested accession is a false positive — GSE52334 is a nomenclature example, not the paper's analysis data; there is no dataset to run a tool over.
  4. Per the 80/20 rule + "honest 1:1": running the pipeline today would only yield a fresh 2026 snapshot, which is explicitly not a reproduction of the 2019 reported numbers. Producing such a number and grading it would be misleading, so it is not attempted.

Evidence captured

Figures / tables: figure
C1
Reported
>30,000 GEO entries missing in ArrayExpress
Reproduced
not attempted
partial
C2
Reported
>10,000 ArrayExpress entries missing in GEO
Reproduced
not attempted
partial
C3
Reported
524 items for keyword 'hypoxia'
Reproduced
not attempted
partial
C4
Reported
58 records (Homo sapiens + Illumina)
Reproduced
not attempted
partial
C5
Reported
95,334 visits (Jul2015-Oct2019, awstats)
Reproduced
non-derivable (private server logs)
did not match
C6
Reported
393,174 page views (awstats)
Reproduced
non-derivable (private server logs)
did not match
C7
Reported
630,837 hits (awstats)
Reproduced
non-derivable (private server logs)
did not match
C8
Reported
GSE52334 corresponds to E-GEOD-52334
Reproduced
true by construction (naming rule)
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 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)
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 🟡
Minor / cosmetic deviation
+1 pts
From: Q6 · Severity of the deviation 🟡
Input / endpoint not comparable 1:1
+1 pts
From: Q1 · Data identity 🔴
Concordant (toward reproduced)
Code + data deposited & functional
-2 pts
From: Data & code availability Available & functional
Total score +7

This is a correctly-handled DROP (no_expected_result): AOE is a database/index resource paper whose every reported number is either a time-varying live-DB snapshot (>30,000 / >10,000 missing entries, 524 'hypoxia' hits, 58 filtered records) or a private awstats web-usage figure (95,334 visits / 393,174 page views / 630,837 hits), with no shipped reference output to reproduce 1:1. The blocker is on the data-availability/input side — the legacy ArrayExpress API is retired (now BioStudies) and the usage logs are private — not on any core-computation or authors'-integrity failure. The only checkable item, the C8 nomenclature mapping GSE52334↔E-GEOD-52334, is true by construction. No fabrication concern: non-derivability here is the normal nature of a live-resource paper, so overall criticality is yellow rather than red.

🤝
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.

58.6 k
tokens (I/O) · 2.8 M incl. cache
6 min
runtime
Per-job HPC accounting not captured for this run — the runtime shown is the reproduction’s measured wall-clock time.