Screening of core genes prognostic for sepsis and construction of a ceRNA regulatory network.
The main results reproduced: recomputed values matched the published ones within tolerance.
Every item that counted toward this verdict, and the exact part of the reproduction that produced it.
- ✓Same input data as the authors
- ✓Reported values were directly comparable
- ✓No relevant deviation in data/preprocessing
- ✓No authors-side cause for any deviation
- ✓Reported values are derivable from the shared data
- ✓Any deviation was negligible
- ✓The central claim held under reproduction
- ✓Overall, the reproduction was clean
- Every checked point held up.
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 for the validation step -> 1:1 (directional). The paper's central biological claim is that 4 core genes (CD247, IL2RB, TGFBR3, IL1R2) are dysregulated in sepsis -- 3 down, IL1R2 up. Using P16 (an existing third-party tool, GEOquery+limma, on the paper's OWN cited public GEO microarray cohorts) we reproduced all 4 directions EXACTLY and with strong statistical support in TWO independent cohorts: GSE28750 (10 sepsis vs 20 healthy, GPL570) and GSE95233 (51 septic-shock admission-day vs 22 healthy, GPL570). Every direction matches (adjP as low as 4.9e-34). The comparison is directional, matching the paper's own validation logic (RT-qPCR effect signs + meta forest plot); we did not attempt to match exact qPCR/meta effect sizes. NOT ATTEMPTED (the hard ~20%, by design): (a) survival of the 4 genes in GSE65682 (optional R3); (b) the headline RNA-seq DE counts 1044/66/155 -- these derive from the authors' own RNA-seq deposited at CNGBdb CNP0002611 (non-GEO raw FASTQ) processed via SOAPnuke->DESeq2 + the iDEP93 web platform, not scriptably pinnable; (c) the ceRNA network (4 mRNA/10 miRNA/23 lncRNA) built from miRWalk+miRDB+OmicShare web tools. No fabrication signal on the checked claims; the unchecked numbers were out of scope, not flagged. Env: R 4.5.3, GEOquery 2.78.0, limma 3.66.0 (conda prefix env on «infra»).
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.
-
v1 current initial assessment Score 86assessed: 2026-06-15 ⛓ 1bc8d7ff0969
✎ 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.
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👤 1 human curator(s) · Level L2 2026-06-15
- 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: sonnetWhether core genes derived from RNA-seq and network analysis of peripheral blood from sepsis patients versus healthy controls can serve as prognostic biomarkers for sepsis, and whether a competing endogenous RNA (ceRNA) regulatory network involving these genes can be constructed.
- ★ RNA-seq of peripheral blood from 23 sepsis patients and 10 healthy controls identifies 1,044 DEmRNAs, 66 DEmiRNAs and 155 DElncRNAs. finding
- ★ DEmRNAs are mainly involved in inflammatory response, immune regulation and neutrophil activation (GO/GSEA). finding
- ★ WGCNA identifies CD247, IL-2Rβ, TGF-βR3 and IL-1R2 as core genes in sepsis. finding
- ★ In an LPS-induced THP-1 sepsis cell model, IL-1R2 is up-regulated while CD247, IL-2Rβ and TGF-βR3 are down-regulated. finding
- ★ CD247, IL-2Rβ and TGF-βR3 expression is positively associated with 28-day survival in sepsis, while IL-1R2 is negatively associated. finding
- ★ A ceRNA regulatory network (10 miRNAs and 23 lncRNAs linked to the 4 core genes) was constructed based on correlation analysis and molecular interaction prediction. resource
- Meta-analysis across multiple GEO microarray datasets shows IL-1R2 up-regulated and IL-2Rβ/TGF-βR3 down-regulated in sepsis vs normal. finding
- WGCNA is used as a method to identify functionally correlated gene modules that predict novel gene function. method
| Assay | System | Perturbation | Readout | Platform |
|---|---|---|---|---|
| RNA-seq (mRNA/miRNA/lncRNA) | peripheral blood, sepsis patients (n=23) vs healthy controls (n=10) | none (disease state: sepsis) | differential gene expression (|FC|>=4, FDR<0.01) | — |
| GO annotation and GSEA functional clustering | DEmRNAs from peripheral blood samples | none | functional/pathway enrichment | R (v4.0.5 for GO, v3.6.3 for GSEA) |
| WGCNA (weighted gene co-expression network analysis) | peripheral blood mRNA expression data | none | gene co-expression modules and core gene identification | iDEP93 online platform |
| Meta-analysis of microarray datasets | blood, Homo sapiens (multiple GEO datasets: GSE28750, GSE54514, GSE95233, GSE6535, GSE63042, GSE74224, GSE67652, GSE12624) | none (sepsis vs normal, sepsis vs SIRS) | expression pattern of core genes | R package "meta" |
| Survival analysis | peripheral blood, sepsis patients, GSE65682 dataset (n=478) | none | 28-day survival association with core gene expression | GraphPad Prism 7 |
| Cell culture / LPS-induced sepsis modeling | THP-1 human monocytic leukemia cell line differentiated to macrophages | PMA differentiation + LPS (100 ng/ml, 6h) to induce sepsis | cellular sepsis model for downstream expression analysis | — |
| RT-qPCR | THP-1-derived macrophages (LPS-induced sepsis model) | LPS-induced sepsis vs untreated control | mRNA expression of CD247, IL1R2, IL2RB, TGFBR3 (2^-ΔΔCt) | SYBR Green kit (TOYOBO, QPK-201) |
- – 1,044 DEmRNAs identified (688 up-regulated, 356 down-regulated) |FC|>=4, FDR<0.01
- – 66 DEmiRNAs identified (29 up-regulated, 37 down-regulated) |FC|>=4, FDR<0.01
- – 155 DElncRNAs identified (61 up-regulated, 94 down-regulated) |FC|>=4, FDR<0.01
- – WGCNA identified CD247, IL-2Rβ, TGF-βR3, IL-1R2 as core genes located centrally in Blue and Green co-expression modules
- – RT-qPCR in THP-1 sepsis model: IL-1R2 up-regulated, CD247/IL-2Rβ/TGF-βR3 down-regulated IL1R2 4.028-fold (log2); CD247 -0.805; IL2RB -0.514; TGFBR3 -0.835
- – Survival analysis: CD247, IL-2Rβ, TGF-βR3 positively correlated with 28-day survival; IL-1R2 negatively correlated
- – Meta-analysis sepsis vs normal: IL-1R2 up, IL-2Rβ and TGF-βR3 down, CD247 no significant difference
- – ceRNA network constructed with 10 miRNAs and 23 lncRNAs correlated with the 4 core genes 10 miRNAs, 23 lncRNAs
- pvalue p<0.0001 (SOFA score, sepsis (n=23) vs control (n=10))
- fold_change |FC|>=4, FDR<0.01 (DER (DEmRNA/DEmiRNA/DElncRNA) screening threshold)
- pvalue t=-4.323, p=0.012 (RT-qPCR CD247 down-regulation in THP-1 sepsis model)
- pvalue t=-4.102, p=0.015 (RT-qPCR IL2RB down-regulation in THP-1 sepsis model)
- pvalue t=-5.120, p=0.007 (RT-qPCR TGFBR3 down-regulation in THP-1 sepsis model)
- pvalue t=9.950, p=0.001 (RT-qPCR IL1R2 up-regulation (4.028 [2.904-5.153]) in THP-1 sepsis model)
- count 23 sepsis, 10 healthy controls (subjects enrolled for RNA-seq/clinical characterization)
- count 478 peripheral blood samples (GSE65682 dataset used for survival analysis)
Statistical methods review
Model: sonnetA neutral, descriptive read of the statistical approach — what was done, and (for shared learning, not as criticism) what could also have been done.
This case-control transcriptomic study compared peripheral blood RNA profiles from 23 sepsis patients and 10 healthy controls using RNA-seq, applying DESeq2 to identify differentially expressed mRNAs, miRNAs, and lncRNAs, followed by WGCNA co-expression analysis to identify four core prognostic genes. Those genes were externally validated by meta-analysis across eight GEO microarray datasets (using fixed- or random-effects models depending on heterogeneity) and by log-rank survival analysis in the GSE65682 cohort. RT-qPCR in an LPS-stimulated THP-1 cell model provided experimental confirmation, with results analyzed by independent-sample t-test and reported as mean difference with 95% CI.
| Test | Applied to | n | Assumptions |
|---|---|---|---|
| DESeq2 Wald test (negative-binomial generalized linear model) | Differential expression of mRNA, miRNA, and lncRNA (sepsis vs. healthy control); thresholds |FC| ≥ 4 and FDR < 0.01 | 23 sepsis patients vs. 10 healthy controls | not stated |
| GO enrichment (hypergeometric test with p-value adjustment, R 4.0.5) | Functional annotation of 1,044 DEmRNAs; significance threshold p < 0.05 | 1,044 DEmRNAs | not stated |
| Gene Set Enrichment Analysis (GSEA, R 3.6.3) | Pathway-level functional clustering of DEmRNAs; thresholds FDR < 0.25 and p.adjust < 0.05 | 1,044 DEmRNAs | not stated |
| WGCNA (weighted gene co-expression network analysis; soft-threshold power = 12, module size > 50) | Co-expression module detection and hub-gene (core gene) identification | 23 sepsis patients vs. 10 healthy controls | not stated |
| Fixed-effects or random-effects meta-analysis (R 'meta' package; model selected by heterogeneity p-value < 0.05) | Expression of CD247, IL-1R2, IL-2Rβ, TGF-βR3 across GEO datasets: sepsis vs. normal (5 datasets) and sepsis vs. SIRS (5 datasets) | Multiple GEO datasets totalling several hundred samples per comparison (individual dataset sizes in Table 2; aggregate N not pooled in the paper) | not stated |
| Log-rank test | 28-day survival analysis for each of the four core genes (high vs. low expression dichotomy), GSE65682; threshold p < 0.05 | 478 sepsis patients (GSE65682, as stated in Methods; Table 2 lists 802 total samples for that dataset) | not stated |
| Independent-sample t-test (two-sample) | RT-qPCR comparison of core gene expression (2^-ΔΔCt) in LPS-stimulated THP-1 cells vs. control | Not stated (number of biological or technical replicates per condition not reported) | stated |
| Independent-sample t-test (two-sample) | Demographic/clinical characteristic comparisons between sepsis and control groups (Table 1) | 23 sepsis patients vs. 10 healthy controls | stated |
| Pearson or Spearman correlation (OmicShare cloud platform; p < 0.05) | ceRNA network construction: pairwise correlation between DEmRNAs, DEmiRNAs, and DElncRNAs to select candidates for the lncRNA-miRNA-mRNA network | 23 sepsis patients vs. 10 healthy controls | not stated |
-
Four separate log-rank tests were performed (one per core gene) for 28-day survival without a stated correction for multiple comparisons↳ Could also: A Bonferroni or Holm correction across the four tests could also be applied; alternatively, a multivariate Cox proportional-hazards model incorporating all four genes simultaneously would also characterize independent prognostic contributions — Adjusting for four simultaneous comparisons controls the family-wise error rate; a Cox model additionally yields hazard ratios with CIs, allows adjustment for clinical covariates such as SOFA score, and avoids the information loss of expression dichotomization
-
Survival was analyzed by dichotomizing expression into 'high' vs. 'low' groups at an unstated threshold before applying the log-rank test↳ Could also: Cox proportional-hazards regression treating gene expression as a continuous predictor would also characterize the survival association without requiring an arbitrary cut-point — Continuous-variable Cox regression avoids information loss inherent in dichotomization and produces a hazard ratio with CI as the effect estimate; the cut-point choice can substantially influence log-rank results in some datasets
-
Four independent-sample t-tests were conducted for RT-qPCR validation of the four core genes without a stated multiplicity correction↳ Could also: Applying a Bonferroni or Holm correction across the four simultaneous comparisons would also control the family-wise error rate — When several tests are performed on the same experimental unit, adjusting for multiplicity limits the probability of at least one false positive across the family; the adjustment is especially straightforward when the family size is small and known in advance
-
The number of biological and technical replicates for the RT-qPCR experiment in THP-1 cells was not reported↳ Could also: Reporting the number of independent biological replicates (distinct cell passages or wells treated independently) separately from technical replicates would also fully characterize the experimental unit — Distinguishing biological from technical replication informs the denominator of the t-test and the generalizability of results; when n per group is very small, a non-parametric Mann-Whitney U test requires no distributional assumption and may also be considered
-
The meta-analysis model (fixed-effects vs. random-effects) was selected empirically based on whether the between-study heterogeneity p-value was < 0.05↳ Could also: Using the random-effects model by default (pre-specified, irrespective of the heterogeneity test result) or reporting results under both models would also be a standard approach — The heterogeneity test has low power when the number of studies is small, so a non-significant result does not confirm homogeneity; pre-specifying the random-effects model avoids a data-driven model choice and is generally more conservative when cross-platform, cross-population differences cannot be excluded
-
Hub (core) genes were selected from WGCNA modules by their visual centrality among the top-80 genes without reporting quantitative module-membership or gene-significance thresholds↳ Could also: Reporting module membership (kME) and gene significance statistics — both standard WGCNA outputs — and applying explicit numerical thresholds (e.g., kME > 0.8, GS > 0.5) for hub-gene selection would also be a standard approach — Explicit, quantitative selection criteria make the hub-gene identification step reproducible and allow readers to assess how centrally each candidate gene sits within its module relative to others
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.
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.
What was reproduced
The exact results taken into scope, with each reported value next to the value our attempt produced.
Scope — PMID 36855106
Title: Screening of core genes prognostic for sepsis and construction of a ceRNA regulatory network. Wang et al., BMC Medical Genomics 2023; 16:37. PMCID PMC9976425 · DOI 10.1186/s12920-023-01460-8.
Paper structure (what the paper actually does)
- Primary DE analysis (authors' own RNA-seq). Peripheral-blood total RNA from
23 sepsis patients + 10 healthy controls, sequenced for lncRNA/miRNA/mRNA.
Raw data deposited at China National GeneBank DataBase (CNGBdb), accession
CNP0002611 — not in GEO. Reads cleaned with SOAPnuke (the repo link in the
brief =
BGI-flexlab/SOAPnuke, a FASTQ-cleaning tool, consistent with a BGI/CNGB-deposited dataset). DE called with DESeq2 at |FC| ≥ 4 & FDR < 0.01 → 1044 DEmRNAs (688 up / 356 down), 66 DEmiRNAs (29/37), 155 DElncRNAs (61/94). Normalization/QC via the iDEP93 web platform. - Core-gene selection. WGCNA on the iDEP93 platform (soft threshold 12, module size > 50); Blue + Green modules correlate with phenotype; top-80 central genes → 4 core genes: CD247, IL2RB (IL-2Rβ), TGFBR3 (TGF-βR3), IL1R2 (IL-1R2).
- Cross-validation of the 4 core genes in public GEO microarray cohorts via a
meta-analysis (R
metapackage) of expression direction:- sepsis-vs-normal: GSE28750, GSE54514, GSE69528, GSE95233, GSE67652
- sepsis-vs-SIRS: GSE28750, GSE6535, GSE63042, GSE74224, GSE12624
- Survival of the 4 genes in GSE65682 (478 sepsis, 28-day survival): CD247, IL2RB, TGFBR3 positively associated with survival; IL1R2 negatively (p < 0.05).
- ceRNA network (lncRNA–miRNA–mRNA): miRWalk (mRNA 3′UTR) + miRDB (DElncRNA) binding + OmicShare correlation → 4 mRNA / 10 miRNA / 23 lncRNA.
- Wet-lab: RT-qPCR on clinical samples + THP-1/LPS in-vitro (out of scope).
In scope (clearly-specified, fully-public, low-hanging — the 80)
Reproduce the paper's central biological claim — that the 4 core genes are dysregulated in the reported directions in sepsis — using the named third-party tools (GEOquery + limma) on the paper's own cited public GEO data (brief rule 2 / P16). Concretely, per public cohort:
- R1 (primary, pinned dataset GSE28750, GPL570): limma DE, sepsis (n=10) vs healthy (n=20). Extract logFC / t / adj.P for CD247, IL2RB, TGFBR3, IL1R2 and test whether direction matches the paper (CD247/IL2RB/TGFBR3 down, IL1R2 up in sepsis).
- R2 (replication, GSE95233, GPL570): same 4-gene direction, septic shock (D0) vs healthy.
- R3 (optional, GSE65682, GPL13667): survival (Cox/log-rank) of the 4 genes vs 28-day mortality — direction (protective vs risk) only.
Pipeline per result: GEOquery (series matrix) → log2 + probe→symbol collapse →
limma lmFit/eBayes (R1/R2) or survival (R3). Tool = limma/GEOquery, exactly the
class of tool the paper used for the validation step.
Out of scope / NOT attempted (the hard ~20 %, with reasons)
- The headline DE counts (1044/66/155) and the WGCNA core-gene derivation. These depend on the authors' own RNA-seq at CNGBdb CNP0002611 (non-GEO repository, raw FASTQ, requires SOAPnuke→align→count→DESeq2) and two web platforms (iDEP93 for normalization+WGCNA, OmicShare for the network) whose exact internal parameters are not scriptable/pinnable. Reproducing these 1:1 is the expensive 20 % and is explicitly skipped per the 80/20 rule. We record CNP0002611 reachability as a data-availability note but do not run the full RNA-seq pipeline.
- ceRNA network (4/10/23). Built from the above DE lists through miRWalk + miRDB + OmicShare web tools — not reproducible without the CNGBdb DE lists and the web DBs' exact versions. Counts recorded as reported claims only.
- RT-qPCR + THP-1/LPS in-vitro — wet-lab, non-pipeline, out of scope.
Possible-fabrication / provenance flags (for the human auditor)
- The Methods name DESeq2 (a count-based R
Assessments & scoring basis
Each contributor’s verdict, the per-question basis, and the auditable, itemised worksheet behind it.
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.
Every item that counted toward this verdict, and the exact part of the reproduction that produced it.
The paper's core biological claim — CD247, IL2RB, TGFBR3 down and IL1R2 up in sepsis — reproduces exactly by direction in two independent public GEO cohorts (adjP as low as 4.9e-34), with no fabrication signal on any checked claim. The only deviations are on our/method side: the comparison is directional (limma logFC vs the paper's qPCR/meta effect sizes) and uses self-defined sample subsets. The paper's titular prognostic survival (C6-C9) and ceRNA network (C10-C12) claims were not attempted because they rest on non-public CNGBdb RNA-seq + web-only tools (data-availability limitation, not an authors' defect). Net: a solid, well-supported reproduction of what was tested, downgraded to yellow overall only for directional-only scope and the substantial untested headline portion.
Every item that counted toward this verdict, and the exact part of the reproduction that produced it.
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-4-8Measured resources invested to assess this paper — sanitised (machine class only, no job ids/paths). Compute = HPC accounting (SLURM); tokens = the AI agent's session.