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

Weighted gene co-expression network analysis reveals that CXCL10, IRF7, MX1, RSAD2, and STAT1 are related to the chronic stage of spinal cord injury.

Ann Transl Med · 2021
44/100 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

The main result did not reproduce in this reproduction attempt. Where our recomputation produced values that differ from the published ones, those discrepancies are listed below. This is a single automated attempt — not peer review and not a finding of error or misconduct — and differences can also arise from data access, undocumented parameters or the computing environment. The verdict can be contested via “report an error”.

Reproduced on the brainbox compute brainarbeit.com
How its reproducibility compares
44/100
Reproducibility score
1.7 SD below mean
vs. all fields · 1173 studies
🎯 Scores higher than 5% of all assessed papers rank 1109 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

INTERIM (re-run in progress, «job»). PARTIAL reproduction. Paper ships no own code; pipeline is the standard WGCNA CRAN tool on public GEO GSE45006 (rat SCI, GPL1355, 24 samples) -> per BRIEF P16 a valid reproduction, NOT a no_code drop. REPRODUCED: (C1) probe->gene collapse gives 14604 genes ~ paper's 14402 (+1.4%); (C7, strongest) the 5 ISG hub genes CXCL10/IRF7/MX1/RSAD2/STAT1 robustly co-cluster in one injury-associated module in every run, eigengene low in sham and elevated 1wk-8wk post-injury; (C2) beta=18 adopted. NOT REPRODUCED: (C3/C4) module count/sizes differ (WGCNA colors are arbitrary size-rank labels); (C5) the headline -0.79 hub-module<->chronic correlation -- we get |r|=0.33-0.52 with biologically-coherent POSITIVE sign (ISGs up in chronic, matching paper's own qPCR), conflicting with the reported NEGATIVE -0.79 (possible internal inconsistency / not derivable as stated); (C6) PPI not comparable. NOT ATTEMPTED: C8 GSE2599 Venn, wet-lab. This file will be refined once re-run jobs complete and outputs are pulled back.

💻 Code ↗ 🗄 Data: GSE45006

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 44
    assessed: 2026-06-16 ⛓ dc178ba32fa9
✎ 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
2026-06-22
Rubric version
not recorded
Assessed by
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

Weighted gene co-expression network analysis (WGCNA) can identify gene co-expression modules and hub genes specifically associated with the chronic stage of spinal cord injury (SCI), revealing molecular mechanisms underlying this stage.

Core claims
  • The brown co-expression module (775 genes) is the module most significantly associated with the chronic stage of SCI finding
  • CXCL10, IRF7, MX1, RSAD2, and STAT1 are hub genes of the brown module identified via PPI network degree analysis and Venn diagram overlap between GSE45006 and GSE2599 finding
  • The five hub genes are significantly upregulated in the chronic phase of SCI compared to normal spinal cord, confirmed by qRT-PCR finding
  • Immune infiltration by CD8+ T cells, macrophages, neutrophils, plasmacytoid dendritic cells, helper T cells, Th2 cells, and tumor-infiltrating lymphocytes may be involved in SCI finding
  • Hub genes are positively correlated with specific infiltrating immune cell types/pathways (CXCL10 and STAT1 with inflammation promotion, IRF7 and RSAD2 with type I IFN response, MX1 with plasmacytoid dendritic cells) finding
  • CD4+ T cells and CD19+ B cells are significantly increased at the spinal cord injury epicenter in SCI rats finding
  • WGCNA on GSE45006 identified 14,402 filtered genes organized into 7 co-expression modules method
  • Brown module genes are functionally enriched in extracellular structure organization, lysosome, isoprenoid biosynthesis, response to wounding, and cholesterol biosynthesis pathways, among others finding
Experimental setups
Assay System Perturbation Readout Platform
Microarray gene expression profiling / WGCNA Rat spinal cord tissue (sham, 1d/3d/1wk/2wk/8wk post-SCI, n=4/group) SCI (aneurysm clip compression/contusion model) Genome-wide gene expression, co-expression modules GEO dataset GSE45006, WGCNA R package
Microarray gene expression profiling (validation cohort) Rat spinal cord tissue (35 days post-SCI, n=3, vs uninjured control, n=3) SCI vs uninjured control Differentially expressed genes GEO dataset GSE2599
Functional enrichment analysis (GO/KEGG) Brown module gene set (in silico) none Enriched biological processes, cellular components, molecular functions, pathways Metascape database
Protein-protein interaction network analysis Brown module genes (top 400, in silico) none Gene connectivity degree, hub gene identification STRING database v11.0, Cytoscape v3.7.2
Quantitative real-time PCR (qRT-PCR) Rat spinal cord tissue (chronic phase SCI vs normal, n=3 pairs) SCI (contusion model via NYU Impactor) mRNA expression of CXCL10, IRF7, MX1, RSAD2, STAT1 SYBR Green (Takara), thermal cycler
Correlation analysis of hub genes with immune infiltration Rat SCI samples (bioinformatic/statistical analysis) SCI Correlation coefficients between hub gene expression and immune cell infiltration scores ggstatsplot R package
Flow cytometry Rat spinal cord injury epicenter tissue SCI CD4+ and CD19+ lymphocyte counts BD Biosciences flow cytometer
Key results
  • Brown module correlates strongly with chronic stage of SCI r=-0.79, P=5E-06
  • Module membership correlates with gene significance for SCI stage in brown module cor=0.6, P=5.8E-77
  • CXCL10, IRF7, MX1, RSAD2, STAT1 mRNA upregulated in chronic SCI vs normal spinal cord by qRT-PCR P<0.05
  • CXCL10 positively correlated with promotion of inflammation cor=0.953913043, P=1.78E-06
  • IRF7 positively correlated with type I IFN response cor=0.968695652, P=1.43E-06
  • RSAD2 positively correlated with type I IFN response cor=0.972173913, P=1.34E-06
  • MX1 positively correlated with plasmacytoid dendritic cells cor=0.931304348, P=2.29E-06
  • CD4+ T cells and CD19+ B cells significantly increased at injury epicenter in SCI rats P<0.001
Key statistics
  • correlation r=-0.79, P=5E-06 (brown module vs chronic stage of SCI trait association)
  • correlation cor=0.6, P=5.8E-77 (brown module membership vs gene significance for SCI stage)
  • correlation cor=0.953913043, P=1.78E-06 (CXCL10 vs promotion of inflammation)
  • correlation cor=0.968695652, P=1.43E-06 (IRF7 vs type I IFN response)
  • correlation cor=0.972173913, P=1.34E-06 (RSAD2 vs type I IFN response)
  • correlation cor=0.931304348, P=2.29E-06 (MX1 vs plasmacytoid dendritic cells)
  • correlation cor=0.951304348, P=1.85E-06 (STAT1 vs promotion of inflammation)
  • count 14,402 genes; 7 modules (genes retained after filtering for WGCNA network construction)

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 study applied WGCNA to two rat spinal cord injury microarray datasets (GSE45006: n=24 samples across six timepoints; GSE2599: n=6 samples) to identify co-expression modules correlated with the chronic injury stage, selecting the brown module (r=−0.79 with 8-week timepoint) and five hub genes by PPI network degree and Venn diagram overlap. Hub genes were validated by qRT-PCR in a rat SCI model (n=3 chronic SCI vs. n=3 controls), and hub gene–immune cell associations were assessed using Pearson r via ggstatsplot. Inferential statistics included classical t-tests for bioinformatic group comparisons, one-way ANOVA with Bonferroni correction for multi-group in vivo comparisons, and repeated-measures two-way ANOVA with Tukey post hoc for behavioral data; all results were reported as mean ± SEM with P<0.05 as the significance threshold.

Replicationbiological Sample sizen=4 per group for GSE45006 (6 groups, 24 total); n=3 per group for GSE2599 (2 groups, 6 total); 30 rats (250±25 g, 10-week-old female Wistar) for in vivo study with n=3 per group for qRT-PCR validation (three pairs stated in Figure 11 caption); no formal power calculation reported GroupsSham/uninjured control vs. SCI at multiple timepoints (1d, 3d, 1wk, 2wk, 8wk) in microarray; chronic SCI vs. uninjured control in qRT-PCR and flow cytometry validation Pairingunpaired Randomization/blindingnot stated DispersionSEM Exact p-valuesyes Effect sizesno Confidence intervalsno Multiplicity correctionBonferroni correction (one-way ANOVA post hoc); Tukey HSD (two-way ANOVA post hoc); P≤0.01 threshold for GO/KEGG enrichment (no named correction method stated); no correction stated for module-trait or hub gene–immune cell Pearson correlations
Statistical tests used
Test Applied to n Assumptions
Classical t-test Hub gene expression comparisons between groups in GSE45006 dataset n=4 per group (24 samples total across 6 groups in GSE45006) not stated
One-way ANOVA with Bonferroni post hoc correction Comparisons among multiple groups (stated in Statistical Analysis section; specific figure not named) not stated
Repeated-measures two-way ANOVA with Tukey post hoc test Behavioral results not stated
Pearson correlation Module-trait associations (brown module vs. chronic stage: r=−0.79, P=5E-06); module membership vs. gene significance (cor=0.6, P=5.8E-77); hub gene vs. immune infiltration scores (five correlations, cor range 0.931–0.973) 24 samples for module-trait correlation; not explicitly stated for immune infiltration correlations not stated
GO and KEGG pathway enrichment analysis via Metascape (P≤0.01 threshold, minimum overlap ≥3) Functional annotation of 775 brown module genes 775 genes in brown module na
Unstated inferential test (P<0.001 reported, marked ***) CD4+ T cell and CD19+ B cell counts by flow cytometry in SCI vs. control rat spinal cord not stated
Approaches that could also have been used
  • Data dispersion was reported as mean ± SEM throughout, including for qRT-PCR validation with n=3 per group
    Could also: Report mean ± SD or 95% confidence intervals — SD directly describes the spread of the observed sample values; 95% CIs convey both variability and the precision of the mean estimate in a single interval and are increasingly recommended by reporting guidelines (e.g., Nature Methods, APA) for small-sample biological data
  • Classical t-tests were used to compare hub gene expression between groups in the microarray data with n=4 per group
    Could also: Apply a moderated t-test via R/limma, or a non-parametric Mann-Whitney U test — Limma's empirical Bayes moderated t-statistic borrows variance information across all genes, stabilizing estimates at very small sample sizes common in microarray studies; Mann-Whitney U is a standard non-parametric alternative that does not require normality, an assumption that cannot be well assessed with n=4
  • Bonferroni correction was applied for one-way ANOVA post hoc comparisons
    Could also: Apply Benjamini-Hochberg false discovery rate (FDR) correction — Bonferroni controls the family-wise error rate conservatively by dividing alpha by the number of tests (α/m); BH-FDR controls the expected proportion of false discoveries among all rejected hypotheses and is widely used as an alternative with higher statistical power when multiple comparisons are made simultaneously
  • Hub genes were ranked and selected by degree (number of direct PPI connections) in the STRING network
    Could also: Also compute betweenness centrality, closeness centrality, or eigenvector centrality as complementary hub-ranking metrics — Degree captures direct connectivity, while betweenness centrality identifies genes that act as bridges between network communities and closeness centrality highlights genes with short topological distances to all others; using multiple centrality measures together can surface hub candidates with distinct network roles that degree alone would not rank highly
  • Multiple Pearson correlations between hub genes and immune cell infiltration scores were reported without a stated correction for multiple comparisons
    Could also: Apply FDR correction (e.g., Benjamini-Hochberg) across the family of correlations, or use Spearman rank correlation — FDR correction accounts for the increased chance of false positives when testing many correlations simultaneously; Spearman rank correlation is a standard non-parametric alternative that does not assume bivariate normality and is robust to outliers, which may be relevant for immune infiltration scores estimated computationally from microarray data
  • The WGCNA soft-power threshold β=18 was selected at a scale-free topology fit index of R²=0.55
    Could also: Select β at a scale-free fit R²≥0.80 as recommended in WGCNA package documentation, or use a signed or signed-hybrid network construction — The WGCNA authors recommend R²≥0.80 as the threshold for adequate scale-free approximation; signed networks additionally preserve the direction of co-expression relationships, which can distinguish co-upregulated from co-downregulated gene sets within the same module and yield more biologically interpretable results
Software: R/WGCNA · Metascape · STRING-db 11.0 · Cytoscape 3.7.2 · R/ggplot2 · R/ggstatsplot · GraphPad Prism 6.0

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
9
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.

What was reproduced

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

Scope — pmid-34532385

Paper: Weighted gene co-expression network analysis reveals that CXCL10, IRF7, MX1, RSAD2, and STAT1 are related to the chronic stage of spinal cord injury. Ann Transl Med 2021. DOI 10.21037/atm-21-3586 · PMCID PMC8421925.

Pipeline (as described in Methods)

  1. Data: GEO GSE45006 (rat spinal cord injury, Affymetrix Rat Genome 230 2.0, GPL1355). Discovery cohort. Validation cohort GSE2599 (35 d post-SCI vs control).
  2. Preprocessing: series-matrix expression, normalized (boxplot medians aligned).
  3. Gene selection: 14,402 probes analyzed → top 5,000 most-variable genes kept for WGCNA.
  4. WGCNA (R WGCNA package): soft-threshold power β = 18; module detection → 7 modules (turquoise 2003, blue 1285, brown 775, yellow 553, grey 248, green 84, red 52; sum = 5000).
  5. Module–trait: the brown module correlates with the chronic SCI stage (8 weeks): r = −0.79, P = 5e-6; module membership cor = 0.6, P = 5.8e-77.
  6. PPI / hub genes: top 400 brown-module genes → STRING v11.0 → Cytoscape v3.7.2; PPI of 367 interactions among 102 genes; intersect with GSE2599 DEGs (Venn, 153 candidate genes) → final hub genes CXCL10, IRF7, MX1, RSAD2, STAT1.

IN SCOPE (pipeline-derived, this room reproduces)

# Result Pipeline Reproducibility
C1 top 5,000 genes selected from ~14,402 probes variance filter high
C2 soft-threshold power β = 18 WGCNA pickSoftThreshold high (deterministic given input)
C3 7 co-expression modules WGCNA blockwiseModules medium (version-sensitive)
C4 module gene counts (7 numbers, sum 5000) WGCNA medium
C5 brown module ↔ chronic stage r ≈ −0.79 WGCNA module-trait cor medium
C6 5 hub genes lie in the SCI-up / brown module & are high-degree STRING/PPI degree medium

OUT OF SCOPE (wet-lab / manual / not pipeline-derived)

  • qRT-PCR validation of the 5 hub genes (wet-lab).
  • Flow cytometry (CD4+/CD19+) (wet-lab).
  • Immune-cell infiltration correlations (ssGSEA/CIBERSORT-style) — borderline computational but underspecified (cell-set definitions not shipped); not graded.
  • Cytoscape cytoHubba ranking exact order (GUI, manual) — we approximate hub-ness by PPI node degree via STRINGdb instead.

Note on the "Code" pointer (P16)

The auto-mined Code URL github.com/IndrajeetPatil/ggstatsplot is a generic ggplot2 stats-plotting package and is NOT the paper's pipeline (false-positive match). Per BRIEF P16, that does NOT make the paper a drop: the paper applies the standard WGCNA pipeline (the canonical third-party tool) to its own public data (GSE45006) with described parameters. We reproduce by running WGCNA on GSE45006 exactly as described. The prior in-room "drop/no_code" verdict was premature and is superseded.

C1
Reported
~14402 genes (probe->symbol collapse) -> top 5000
Reproduced
31099 probes -> 14604 unique genes (collapseRows MaxMean) -> top 5000
within tolerance
C2
Reported
soft-threshold beta=18 (R^2~0.55)
Reproduced
beta=18 adopted; R^2@18=0.73; pickSoftThreshold auto-estimate degenerate(=1)
partial
C3
Reported
7 modules incl grey (6 non-grey)
Reproduced
8 unsigned / 9 signed incl grey
partial
C4
Reported
module sizes turq2003/blue1285/brown775/yellow553/grey248/green84/red52
Reproduced
turq1465/blue997/brown984/yellow643/grey492/green168/red162/black89 (color labels not gene-for-gene comparable)
did not match
C5
Reported
brown(hub) module vs chronic r=-0.79, P=5e-6
Reproduced
hub-gene module vs chronic r=+0.33/+0.34; vs time +0.44/+0.52 (POSITIVE, |r|<<0.79)
did not match
C6
Reported
PPI 367 interactions among 102 genes
Reproduced
1069 edges / 280 nodes (STRING>=400, mislabeled module; cutoff-sensitive; not comparable)
did not match
C7
Reported
hub genes CXCL10/IRF7/MX1/RSAD2/STAT1 co-clustered, chronic-related
Reproduced
all 5 co-cluster in exactly one module in every run; eigengene lowest in sham, elevated 1wk-8wk post-injury
within tolerance
C8
Reported
153 GSE45006/GSE2599 Venn candidates
Reproduced
not attempted (out of 80/20 core; needs separate GSE2599 DEG analysis)
partial

Assessments & scoring basis

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

No assessment has been recorded yet.
🤝
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.

321.9 k
tokens (I/O) · 26.6 M incl. cache
101 min
runtime · 0.12 CPU-h
12.5 GB
peak RAM
4
HPC jobs
hummel
machine