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

Lipopolysaccharide distinctively alters human microglia transcriptomes to resemble microglia from Alzheimer's disease mouse models.

Dis Model Mech · 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

EXACT 1:1 REPRODUCTION. Ran the authors' own PokeMicro Seurat pipeline (Seurat 4.3.0.1) on the paper's own deposited GSE186301 count matrix on «our HPC». All 9 paper claims + 1 code-comment cross-check reproduced EXACTLY: 20231 cells, 12335 genes (min.cells=100), 8 clusters, fibroblast/proliferating clusters of 469/302 cells, 19460 microglia, and the three headline DEG counts 904 (LPS+IFNg 24h) / 802 (ATPgS 24h) / 152 (PGE2 24h) via FindConservedMarkers+metap Tippett — all to the single gene; integration features=1620 matched too. No discrepancies, no fabrication concerns. NOT attempted (out of scope): wet-lab, Cell Ranger/STAR from raw FASTQ (deposited matrix is the documented start), GO enrichment cosmetics, and the GSE109329 external mouse comparator dataset.

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 100
    assessed: 2026-06-20 ⛓ 06a30517f784
✎ 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
2026-06-20
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

The paper tests whether combinatory inflammatory stimulation (ATPγS, LPS+IFN-γ, PGE2, alone or combined) of human iPSC-derived microglia produces transcriptional changes relevant to modeling Alzheimer's disease, compared against transcriptomic changes in AD mouse models and post-mortem AD patient microglia.

Core claims
  • iPSC-microglia show a shared core transcriptional response to ATPγS and to LPS+IFN-γ, suggesting a convergent mechanism of action finding
  • DEGs from all iPSC-microglia treatment conditions significantly overlap with genes altered in microglia from AD patients, though often with inconsistent direction finding
  • A common transcriptomic disease axis exists across AD genetic mouse models that separates wild-type and transgenic AD mouse microglia finding
  • Only LPS(+IFN-γ) shifts microglial transcriptional profiles along the identified disease axis in both mouse and human iPSC-microglia, making it the best modeled stimulus for AD among those tested finding
  • PGE2 produces a milder, more temporally complex/delayed transcriptional response compared to ATPγS and LPS+IFN-γ finding
  • Combining ATPγS with LPS+IFN-γ or PGE2 produces little additional (synergistic) transcriptional effect beyond ATPγS alone finding
  • DEG protein products form protein-protein interaction networks with more interactions than expected by chance, supporting functional convergence of the ATPγS and LPS+IFN-γ responses finding
  • CITE-seq single-cell transcriptomics was used to simultaneously measure iPSC-microglia responses to multiple stimuli and timepoints method
Experimental setups
Assay System Perturbation Readout Platform
single-cell RNA-seq (CITE-seq) human iPSC-derived microglia ATPγS (1 mM) single-cell transcriptome / differential gene expression CITE-seq
single-cell RNA-seq (CITE-seq) human iPSC-derived microglia LPS+IFN-γ (10 ng/ml) single-cell transcriptome / differential gene expression CITE-seq
single-cell RNA-seq (CITE-seq) human iPSC-derived microglia PGE2 (500 nM) single-cell transcriptome / differential gene expression CITE-seq
single-cell RNA-seq (CITE-seq) human iPSC-derived microglia combined treatment (PGE2 or LPS+IFN-γ for 24 h, then ATPγS added for a further 24 h) single-cell transcriptome / differential gene expression CITE-seq
Ca2+ imaging human iPSC-derived microglia pre-treatment with PGE2 or LPS+IFN-γ followed by ATPγS calcium response magnitude
protein-protein interaction network analysis protein products of iPSC-microglia DEGs none (computational analysis of DEG sets) number of network interactions vs randomized control
transcriptomic meta-analysis (single-cell/snRNA-seq) mouse microglia (wild-type and transgenic AD genetic models) genetic (AD mouse model genotype) common disease axis of transcriptomic change
single-nucleus/single-cell RNA-seq (published datasets) human post-mortem microglia (entorhinal cortex and prefrontal cortex, AD patients vs controls) none (AD disease state vs control) cell-type-specific differentially expressed genes
Key results
  • Largest number of DEGs detected after 24 h LPS+IFN-γ exposure n=904 (combined P<0.05)
  • 24 h ATPγS exposure produced the second-largest DEG set n=802 (combined P<0.05)
  • 24 h PGE2 exposure produced fewer DEGs than ATPγS or LPS+IFN-γ n=152 (combined P<0.05)
  • 73 DEGs overlap across all three treatments at 24 h; strong pairwise overlaps between LPS+IFN-γ and ATPγS n=514 overlap, P≈0 (log(P)=-1007.81)
  • Gene expression fold changes at 24 h correlate strongly between LPS+IFN-γ and ATPγS treatments r=0.625, P<2.2×10^-16
  • Only 20 unique DEGs found for LPS+IFN-γ(48h)+ATPγS(24h) combined vs ATPγS alone; only 2 unique DEGs for PGE2(48h)+ATPγS(24h) vs ATPγS alone 20 and 2 unique DEGs respectively
  • Overlap of AD-associated microglial DEGs between two independent post-mortem AD datasets (Grubman and Mathys) higher than expected by chance 12 overlapping genes, P=4.525×10^-12
  • iPSC-microglia DEGs show higher-than-expected overlap with AD patient microglia DEGs, but with some divergent directionality (e.g., SPP1)
Key statistics
  • correlation r=0.625, P<2.2×10^-16 (Correlation of 24h gene expression fold changes between LPS+IFN-γ and ATPγS treatments)
  • pvalue P≈0, log(P)=-1007.81 (Hypergeometric overlap of DEGs between LPS+IFN-γ and ATPγS at 24h)
  • pvalue P=8.23×10^-62 (Hypergeometric overlap of DEGs between LPS+IFN-γ and PGE2 at 24h)
  • pvalue P=4.51258×10^-101 (Hypergeometric overlap of DEGs between ATPγS and PGE2 at 24h)
  • count 904 DEGs (DEGs after 24h LPS+IFN-γ exposure, combined P<0.05)
  • count 802 DEGs (DEGs after 24h ATPγS exposure, combined P<0.05)
  • count 152 DEGs (DEGs after 24h PGE2 exposure, combined P<0.05)
  • pvalue P=4.525×10^-12 (Overlap of AD microglial DEGs between Grubman et al. and Mathys et al. post-mortem datasets)

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.

The paper used single-cell RNA sequencing (CITE-seq) to compare iPSC-derived microglia treated with various stimuli (ATPγS, LPS+IFN-γ, PGE2, and combinations) at 24h and 48h against untreated controls, across four biological replicates (donors). Differential expression was assessed per gene using a 'combined P-value' integrated across donors, gene set overlaps between conditions/datasets were tested with the hypergeometric test, fold-change relationships were assessed with Pearson correlation, Gene Ontology enrichment was reported with adjusted P-values, and protein-protein interaction network enrichment was assessed via randomization-based empirical P-values controlling for node degree and gene length.

Replicationbiological Sample sizeFour biological replicates (iPSC donors) across eight experimental conditions; single-cell counts reported for specific clusters/analyses (e.g. 20,231 total cells post-demultiplexing, 19,460 microglia-like cells used in PCA, cluster 6 n=469, cluster 7 n=302) GroupsEach stimulus (ATPγS, LPS+IFN-γ, PGE2, and combinations) at 24h and/or 48h vs 0h untreated control, plus cross-comparison to human AD post-mortem microglia datasets and mouse AD model microglia Pairingunclear Randomization/blindingnot stated Dispersionunclear Exact p-valuesyes Effect sizesyes Multiplicity correctionyes
Statistical tests used
Test Applied to n Assumptions
Hypergeometric test Overlap of DEG sets between treatments (Fig. 2A), between AD post-mortem datasets, and between iPSC-microglia DEGs and AD patient microglia DEGs (Fig. 4) Overlap counts among DEG sets as stated (e.g. n=514, n=89, n=112, n=605, n=66) not stated
Combined P-value (integrated across biological replicates/donors) Differential expression analysis between each treatment and 0h control cells, per gene Four biological replicates (donors); DEG counts reported (e.g. n=904 for LPS+IFN-γ 24h, n=802 for ATPγS 24h, n=152 for PGE2 24h) not stated
Pearson correlation Correlation between gene expression fold changes at 24h for LPS+IFN-γ vs ATPγS (Fig. S7) Not explicitly stated (gene-level fold changes) not stated
Gene Ontology enrichment analysis (adjusted P-value) Enriched biological processes among up/downregulated DEGs per treatment and time point (Fig. 2B) DEG lists per condition not stated
Randomization-based empirical P-value test Enrichment of protein-protein interactions among DEG products relative to chance, controlling for node degree and gene length (Fig. S10A) DEG-derived PPI network nodes; P<0.0001 based on randomizations not stated
Approaches that could also have been used
  • Overlap between gene sets (e.g. DEGs across treatments, or DEGs vs AD patient microglia genes) was assessed with the hypergeometric test.
    Could also: A permutation-based or GSEA-style enrichment test could also be used. — Permutation approaches can account for gene-level covariates such as expression level or gene length that a simple hypergeometric test assumes are uniform across the background set.
  • Differential expression across donors was summarized using a 'combined P-value' per gene, integrating replicate information.
    Could also: A mixed-effects/pseudobulk approach (e.g. treating donor as a random effect in a model such as MAST, or aggregating to pseudobulk counts for analysis in DESeq2/edgeR/limma) could also be used. — Since donor-to-donor clustering was observed even among untreated controls, explicitly modeling donor as a random effect can help partition biological replicate variance from treatment effects.
  • Gene-level differential expression significance was thresholded at combined P-value<0.05 without a stated multiple-testing correction across the many genes tested.
    Could also: A false discovery rate correction (e.g. Benjamini-Hochberg) applied across all genes tested could also be reported alongside or instead of raw P-values. — FDR correction is standard practice in transcriptome-wide differential expression analysis to control the expected proportion of false positives when thousands of genes are tested simultaneously.
  • Gene Ontology enrichment results were reported using an adjusted P-value without the specific correction method named in the text.
    Could also: Explicitly stating the correction method (e.g. Benjamini-Hochberg FDR, commonly used by tools such as topGO or clusterProfiler) could also be reported. — Naming the specific multiple-testing correction improves reproducibility and lets readers judge the stringency of the enrichment threshold used.
  • The relationship between fold changes across treatments (e.g. LPS+IFN-γ vs ATPγS) was quantified using Pearson correlation.
    Could also: Spearman's rank correlation could also be used to assess this relationship. — Fold-change distributions from differential expression analyses are often skewed or contain outliers, and a rank-based correlation can capture monotonic relationships without assuming linearity or normality.
  • Enrichment of protein-protein interactions among DEG products was tested using empirical P-values from randomizations controlling for node degree and gene length.
    Could also: Additional network-based methods such as edge-permutation null models or pathway topology-based approaches (e.g. SPIA) could also be used to characterize network-level convergence. — These complementary approaches can further probe whether enrichment reflects specific pathway topology rather than degree/length-matched connectivity alone.

What was reproduced

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

Scope — pmid-36254682

Paper: Monzón-Sandoval et al. 2022, Dis Model Mech — "Lipopolysaccharide distinctively alters human microglia transcriptomes to resemble microglia from Alzheimer's disease mouse models." PMID 36254682 / PMC9612871.

Repo: https://github.com/jmonzon87/PokeMicro @ commit ad29f257428709a1237093d02bc83ee2ca5866e1 (2022-03-09). Authors' own code (R / RMarkdown, Seurat). Two scripts: iPSCmicro.Rmd (clustering + pooled-Wilcoxon DEA) and iPSCmicro_Integrated.Rmd (per-donor integration + FindConservedMarkers DEA — this is the source of the paper's headline DEG counts).

Primary data: GEO GSE186301 — scRNA-seq (10x + CITE-seq HTO) of iPSC-derived microglia, 4 donors × 8 stimulation conditions. Repo starts from the deposited protein-coding count matrix + metadata (NOT raw FASTQ), so Cell Ranger / STAR alignment is out of scope (no FASTQ needed; we begin at the count matrix — this is exactly the pipeline the authors documented).

Secondary data: GEO GSE109329 — external mouse microglia stimulus data used only for a cross-species comparison figure. Out of scope for the computational re-run (comparison/illustrative, not pipeline-derived from our matrix).

In scope (pipeline-derived, attempted)

Starting from GSE186301_Counts_ProteinCodingGenes.txt + GSE186301_Metadata.txt, run the documented Seurat workflow and regenerate these reported numbers:

id reported value paper location pipeline
C1 20,231 single cells Results, opening matrix dim / CreateSeuratObject
C6 12,335 genes after min.cells=100 code comment (iPSCmicro.Rmd L62) CreateSeuratObject
C2 8 clusters Results FindClusters res=0.1, dims 1:20
C3 cluster 6 (fibroblast-like) = 469 cells Results cluster table
C4 cluster 7 (proliferating) = 302 cells Results cluster table
C5 19,460 iPSC-microglia after removing clusters 6,7 Results subset idents 0–5
C7 LPS+IFNγ 24h DEGs = 904 (combined P<0.05) Results integration + FindConservedMarkers (Tippett minimump)
C8 ATPγS 24h DEGs = 802 Results ""
C9 PGE2 24h DEGs = 152 Results ""

Stage A (iPSCmicro.Rmd core): C1, C6, C2, C3, C4, C5. Stage B (iPSCmicro_Integrated.Rmd core): C7, C8, C9 — donor-split integration (SelectIntegrationFeatures → FindIntegrationAnchors → IntegrateData) then FindConservedMarkers per condition vs Control, combine per-donor P with metap::minimump (Tippett), count rows with combined P < 0.05.

Out of scope (not attempted)

  • Wet-lab: iPSC differentiation, CITE-seq hashing, FACS (manual/experimental).
  • Cell Ranger / STAR alignment + HTO demultiplexing (raw FASTQ; deposited matrix is the documented entry point).
  • GO/rrvgo enrichment heatmaps, all PDF figure cosmetics (visual, not a reported scalar; expensive, no headline number).
  • GSE109329 mouse cross-species comparison (illustrative external dataset).

Reproducibility notes / risks

  • Cluster counts + DEG counts are Seurat-version sensitive (method card: Seurat 71% clean — pin version + seed). Using r-seurat 4.3.0 (authors' era was Seurat v4, 2022). Seurat sets deterministic seeds by default (FindClusters random.seed, RunUMAP seed.use=42).
  • Code comment says 19,448 microglia cells while paper text says 19,460 (=20,231−469−302). We grade against the paper's 19,460 and report both.
Figures / tables: Fig1
C1
Reported
20231 single cells
Reproduced
20231
exact
C6
Reported
12335 genes after min.cells=100
Reproduced
12335
exact
C2
Reported
8 clusters
Reproduced
8
exact
C3
Reported
cluster6 fibroblast=469 cells
Reproduced
469
exact
C4
Reported
cluster7 proliferating=302 cells
Reproduced
302
exact
C5
Reported
19460 iPSC-microglia
Reproduced
19460
exact
C7
Reported
904 DEGs LPS+IFNg 24h (combined P<0.05)
Reproduced
904
exact
C8
Reported
802 DEGs ATPgS 24h
Reproduced
802
exact
C9
Reported
152 DEGs PGE2 24h
Reproduced
152
exact
X1
Reported
1620 integration features (code comment)
Reproduced
1620
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.

221.1 k
tokens (I/O) · 16.5 M incl. cache
49 min
runtime · 0.11 CPU-h
10.8 GB
peak RAM
1
HPC jobs
hummel
machine