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

Trem2 promotes anti-inflammatory responses in microglia and is suppressed under pro-inflammatory conditions.

Hum Mol Genet · 2020
L1 67/100 3/4
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: 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 🟡
Concordant (toward reproduced)
Code + data deposited & functional
-2 pts
From: Data & code availability Available & functional
Total score +4
✓ What held up
  • Same input data as the authors
  • Reported values were directly comparable
  • The central claim held under reproduction
What did not (or only partly)
  • 🟡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
  • 🟡Overall, the reproduction showed a material discrepancy
How its reproducibility compares
67/100
Reproducibility score
0.4 SD below mean
vs. all fields · 1173 studies
🎯 Scores higher than 29% of all assessed papers rank 795 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

Described well enough for a faithful PARTIAL reproduction via the third-party-tool route (CoExpNets is the generic WGCNA tool; paper-specific scripts not shipped, so applied the described Salmon->tximport->DESeq2 pipeline to the paper's own GSE157891 data, brief P16). CLEAN 1:1 points: (C1) expressed-gene count 11119->11204 (within-tol, deterministic from shipped TPM); (C5) the paper's central claim, Trem2 suppressed in Trem2-R47H KI microglia, reproduces exactly (log2FC=-1.64, padj=2.5e-4). PARTIAL: the exact DEG tallies (C3 408/311, C4 184/83) come out same-direction but ~0.5-0.6x / within-tens-of-percent (C3 253/149 default, C4 155/53 pooled) because Methods give NO software versions, NO exact DESeq2 contrast, and NO Salmon flags -> counts are sensitive to Ensembl release, salmon options, and pre-filter choice. C2 network-gene count (10463) is underspecified (CoV rule not pinnable; closest 9380). NOT ATTEMPTED (last-20%): WGCNA/CoExpNets module structure (OrangeRed3/Salmon4, hub genes, module-trait r) - stochastic module naming + unspecified soft-power/minModuleSize/k; cross-dataset enrichment p-values (4.70e-7, 1.8e-12) - external reference sets not pinnable; wet-lab assays (non-computational). No value appears fabricated; all reported numbers are plausible outputs of the described pipeline under reasonable parameters.

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 67
    assessed: 2026-06-15 ⛓ e08218e5c7a5
✎ 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

This study tests the role of TREM2 in microglial function and inflammatory polarization by reducing Trem2 expression (via siRNA knockdown or the AD-risk R47H knock-in mutation), hypothesizing that decreased TREM2 impairs microglial survival/phagocytosis and the IL-4-driven anti-inflammatory response while being suppressed under pro-inflammatory conditions.

Core claims
  • Trem2 promotes/drives the IL-4-induced anti-inflammatory gene expression program (e.g. Arg1, Ap1b1) in microglia, and reduced Trem2 attenuates this response. finding
  • LPS pro-inflammatory stimulation suppresses Trem2 expression, preventing TREM2's anti-inflammatory drive. finding
  • Trem2 knockdown decreases STAT6 levels, and Arg1-co-regulated genes are enriched for STAT6 transcription factor recognition elements, implicating STAT6 as a downstream mechanism. mechanism
  • The Trem2 R47H knock-in mouse primarily causes decreased Trem2 expression (via a mouse-specific splice effect), serving as an in vivo model of Trem2 down-regulation. finding
  • Reduced Trem2 expression (R47H KI) decreases microglial density and CD68-positive microglia in the hippocampus. finding
  • Acute Trem2 knockdown by siRNA impairs microglial phagocytosis, validating the in vitro preparation. finding
  • Parallel primary microglial model comparing siRNA Trem2 knockdown with Trem2 R47H KI microglia under LPS/IL-4 polarization. method
  • Reduced Trem2 increases microglial apoptosis/cell death. finding
Experimental setups
Assay System Perturbation Readout Platform
RT-qPCR gene expression hippocampal homogenates from Trem2 R47H KI (HO/HE) and WT mice, 4 months old R47H knock-in point mutation Trem2 and microglial gene expression relative to Rps28
Immunohistochemistry / immunostaining hippocampal CA1 region (SO, SP, SR, SLM layers), WT and HO Trem2 R47H KI mice R47H knock-in mutation microglial density (IBA1/AIF1) and CD68-positive microglia density/proportion anti-IBA1, anti-CD68 antibodies, DAPI
Primary microglial culture + RT-qPCR primary mouse microglia (WT and Trem2 R47H KI) from mixed glial culture LPS pro-inflammatory or IL-4 anti-inflammatory stimulation; Trem2 siRNA knockdown vs non-targeting control pro-inflammatory (Tnf, Il1b) and anti-inflammatory (Arg1, Tgfb1) gene expression
RNA-seq primary microglia IL-4 treatment with/without Trem2 knockdown genome-wide transcriptional program (e.g. Arg1, Ap1b1) and STAT6 motif enrichment
ELISA primary microglial conditioned medium/supernatants (WT and HO Trem2 R47H KI) LPS treatment soluble TREM2 protein levels (also TNF-alpha protein)
Immunocytochemistry primary microglia 24h after isolation none culture purity (% IBA1-positive cells) anti-IBA1 antibody, DAPI
FACS / Annexin V apoptosis assay primary microglia from Trem2 R47H KI vs WT mice R47H knock-in mutation % Annexin V-positive, propidium iodide-negative cells (cell death)
Phagocytosis assay primary microglia Trem2 siRNA knockdown vs non-targeting control uptake of pHrodo-conjugated E. coli pHrodo pH-sensitive fluorescent dye
Key results
  • IL-4-induced Arg1 up-regulation was attenuated by Trem2 siRNA knockdown, significant at 48h post IL-4
  • In Trem2 R47H KI microglia, IL-4-induced Arg1 induction was prevented in a gene dose-dependent manner P=0.01 WT vs homozygous
  • LPS strongly down-regulated Trem2 expression and decreased soluble TREM2 protein in conditioned medium
  • Trem2 knockdown decreased STAT6 levels
  • Homozygous Trem2 R47H KI mice showed decreased total microglia density and decreased CD68-positive microglia density/proportion across all four CA1 layers
  • Trem2 siRNA knockdown reduced microglial phagocytosis of pHrodo-E. coli 48.1 ± 9.1% reduction
  • Trem2 R47H KI microglia showed increased cell death vs WT 14.6% vs 7.3% Annexin V+/PI-
  • Trem2 knockdown down-regulated microglial genes (C1qa, Cd68, Csf1r, Igf1, Pik3cg, Spi1, Tnf, Tgfb1) under non-stimulated conditions
Key statistics
  • other 48.1 ± 9.1% reduction in phagocytosis (Trem2 siRNA knockdown vs non-targeting control microglia)
  • count 14.6% Annexin V-positive/PI-negative (Trem2 R47H KI primary microglia cell death)
  • count 7.3% Annexin V-positive/PI-negative (WT littermate primary microglia cell death)
  • other 97.2 ± 0.3% IBA1-positive cells (primary microglial culture purity 24h after isolation)
  • other ~80% knockdown of Trem2 mRNA (siRNA seq 2); 50–60% (siRNA seq 3) (siRNA knockdown efficiency vs non-targeting control)
  • pvalue P<0.001 (Arg1 difference between Trem2 knockdown and control at 48h post IL-4 (Sidak's))
  • pvalue P=0.05 (two-way ANOVA interaction between Trem2 knockdown and IL-4 treatment time for Arg1)
  • pvalue P=0.002 (one-way ANOVA for Arg1 induction across R47H KI genotypes; P=0.01 WT vs homozygous)

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.

The study combined an in vivo mouse model (Trem2 R47H knock-in vs. wild type littermates) with in vitro primary microglial cultures using siRNA knockdown, reading out gene expression by RT-qPCR plus protein by ELISA, immunohistochemistry-based cell counts, and RNA-seq. Group comparisons were made predominantly with one-way and two-way ANOVA followed by Sidak's post hoc pairwise tests, with a Student's t-test used for a single two-timepoint comparison. Results were reported as mean ± SEM with significance shown as P-value thresholds (asterisk tiers) and sample sizes given as numbers of mice or independent experiments.

Replicationmixed Sample sizeReported as number of mice per group/genotype or number of independent experiments per figure; no formal power/sample-size calculation described GroupsWT vs heterozygous/homozygous Trem2 R47H KI mice; Trem2 siRNA vs non-targeting control; LPS or IL-4 vs basal Pairingunclear Randomization/blindingnot stated DispersionSEM Exact p-valuesyes Effect sizesno Confidence intervalsno Multiplicity correctionSidak's multiple comparisons (post hoc after ANOVA)
Statistical tests used
Test Applied to n Assumptions
One-way ANOVA with Sidak's post hoc Hippocampal gene expression across genotypes (Fig 1A, Trem2 expression) N = 6–7 mice per group not stated
Two-way ANOVA (genotype x CA1 layer) Microglial density, CD68+ density, proportion CD68+ in CA1 (Fig 1C) N = 5–6 mice per group not stated
One-way ANOVA Tnf, Il1b, Tgfb1 expression with LPS/IL-4 treatment (Fig 3B) N = 3–6 independent experiments not stated
Student's t-test Arg1 expression, 24 h vs 48 h IL-4 (Fig 3B; Arg1 undetected in control) not stated
Two-way ANOVA with Sidak's post hoc Trem2 expression (Fig 3C) and soluble TREM2 ELISA (Fig 3D), treatment x genotype N = 6 (3C); N = 3–4 (3D) mice per genotype not stated
Two-way ANOVA (interaction) with Sidak's multiple comparisons Arg1 with Trem2 knockdown x IL-4 time (Fig 4B); one-way ANOVA with Sidak for Arg1 in R47H KI (Fig 5B) N = 6 mice per genotype (Fig 5) not stated
Approaches that could also have been used
  • Dispersion was summarized as mean ± SEM throughout the figures.
    Could also: SD or a 95% confidence interval could also be reported alongside or instead of SEM. — SD conveys the spread of the data directly and a 95% CI conveys precision of the estimate; both are often favored, particularly with small n, for communicating variability.
  • Significance was frequently shown using P-value threshold tiers (asterisks) in addition to some exact P-values.
    Could also: Reporting exact P-values uniformly together with effect-size estimates (e.g. mean differences with CIs) could also be done. — Exact values and effect sizes give readers the magnitude and precision of differences, complementing yes/no significance thresholds.
  • Several individual genes were each tested across genotype/treatment using separate ANOVAs (a battery of microglial genes).
    Could also: A family-wise or false-discovery-rate correction (e.g. Benjamini-Hochberg) across the panel of genes could also be applied. — Correcting across the gene family would control the overall error rate when many genes are screened in parallel.
  • Group comparisons used parametric ANOVA and t-tests.
    Could also: Non-parametric equivalents (Kruskal-Wallis, Mann-Whitney U) or explicit checks of normality/variance could also be used. — With small samples these can be informative when distributional assumptions are uncertain, and stating assumption checks documents the basis for the parametric choice.
  • A Student's t-test compared Arg1 at 24 h vs 48 h of IL-4 because it was undetectable in controls.
    Could also: A repeated-measures/mixed-effects model treating experiment as a random factor could also handle the time course. — Mixed models can account for the paired/nested structure of repeated measures from the same cultures and use all timepoints jointly.
  • RNA-seq was mentioned for the IL-4 gene program alongside RT-qPCR validation.
    Could also: Established count-based pipelines (e.g. DESeq2 or limma-voom) with shrinkage and FDR control could also be specified for the differential-expression analysis. — These methods model count dispersion and provide built-in multiple-testing control suited to genome-wide expression data.

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
215
Impact: very high
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.

RRID:AB_915783 RRID in Article (http://semanticscience.org/resource/SIO_001029)
also used by 2 papers:
RRID:AB_2305186 RRID in Article (http://semanticscience.org/resource/SIO_001029)
also used by 1 paper:
RRID:AB_2315049 RRID in Article (http://semanticscience.org/resource/SIO_001029)
also used by 1 paper:
RRID:AB_322219 RRID in Article (http://semanticscience.org/resource/SIO_001029)
also used by 1 paper:
RRID:AB_839504 RRID in Article (http://semanticscience.org/resource/SIO_001029)
also used by 1 paper:
RRID:AB_10641962 RRID in Article (http://semanticscience.org/resource/SIO_001029)
no other assessed paper uses this yet
RRID:AB_11220421 RRID in Article (http://semanticscience.org/resource/SIO_001029)
no other assessed paper uses this yet
RRID:AB_2208679 RRID in Article (http://semanticscience.org/resource/SIO_001029)
no other assessed paper uses this yet
RRID:AB_2255933 RRID in Article (http://semanticscience.org/resource/SIO_001029)
no other assessed paper uses this yet
RRID:AB_356109 RRID in Article (http://semanticscience.org/resource/SIO_001029)
no other assessed paper uses this yet
RRID:AB_477010 RRID in Article (http://semanticscience.org/resource/SIO_001029)
no other assessed paper uses this yet
RRID:IMSR_JAX:027918 RRID in Article (http://semanticscience.org/resource/SIO_001029)
no other assessed paper uses this yet
rs75932628 RefSNP in Introduction (http://purl.org/orb/Introduction)
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-32959884

Paper: Liu et al. 2020, Trem2 promotes anti-inflammatory responses in microglia and is suppressed under pro-inflammatory conditions. Hum Mol Genet, PMID 32959884, PMCID PMC7689298, DOI 10.1093/hmg/ddaa209.

Named code: https://github.com/juanbot/CoExpNets (Botia lab; "Co-expression network management based on WGCNA + k-means"). Public, Apache-2.0, not archived, last push 2021-05-11. Co-author Botia JA. The repo is the generic WGCNA/k-means tool — it does not contain this paper's specific analysis scripts (DESeq2 calls, the exact module run). Per brief rule P16, reproducing by applying the described pipeline to the paper's own data is equally valid.

Data: GEO GSE157891 (SRA SRP282249 / PRJNA663188). 12 mouse primary-microglia RNA-seq samples, paired-end Illumina HiSeq 2500. 2×2 design, n=3: genotype (WT vs Trem2 R47H KI) × treatment (Ctrl vs IL-4). Shipped processed file: GSE157891_geneQuantification_TPM.csv.gz (gene-level TPM, 48,623 rows × 12 samples). Raw FASTQ on ENA (24 files, ~23 GB).

Sample → group map (from GEO characteristics)

SRR title genotype treatment
SRR12630894 T13_Ctrl WT Ctrl
SRR12630895 T13_IL4 WT IL4
SRR12630896 T15_Ctrl R47H Ctrl
SRR12630897 T15_IL4 R47H IL4
SRR12630898 T22-1 R47H Ctrl
SRR12630899 T22-2 R47H IL4
SRR12630900 T32-1 R47H Ctrl
SRR12630901 T32-2 R47H IL4
SRR12630902 T33-1 WT Ctrl
SRR12630903 T33-2 WT IL4
SRR12630904 T43-1 WT Ctrl
SRR12630905 T43-2 WT IL4

Pipeline-derived results (per Methods)

Quantification: Salmon (ENSEMBL GRCm38 annotation) → tximport → log2 TPM. DE: DESeq2, FDR < 0.05. Pre-filter: gene "expressed" if mean log2 TPM > 1.5. Network: WGCNA (CoExpNets) on genes with CoV > 5%.

ID Result Reported Location In scope? Tier
C1 expressed genes (mean log2 TPM > 1.5) 11,119 Methods (RNA-seq) YES — deterministic from shipped TPM 1 (cheap)
C2 network genes (CoV > 5%) 10,463 Methods (network) YES but CoV definition underspecified 1 (cheap)
C3 DEG R47H vs WT, basal 408 up / 311 down Fig 6A YES — needs Salmon→DESeq2 from FASTQ 2 (heavy)
C4 DEG IL-4 vs Ctrl 184 up / 83 down Fig 6C YES — needs Salmon→DESeq2; exact contrast (WT-only vs pooled) underspecified 2 (heavy)
C5 Trem2 decreased in R47H qualitative (FDR<0.05) Fig 6A / text YES — direction check 2

Out of scope / last-20% (not chased)

  • WGCNA module structure (OrangeRed3 module containing Trem2/Tyrobp/Spi1/Stat6; Salmon4 module for IL-4; hub genes Nckap1l/Cd53/Adam8/Fxyd5; module–trait Pearson r). CoExpNets uses WGCNA+k-means with a stochastic color/number module naming that is not byte-reproducible, and the exact run parameters (soft power, minModuleSize, k) are not given. Noted, not reproduced.
  • Cross-dataset enrichment p-values (e.g. P=4.70e-7 hippocampal-microglia module overlap; Fisher P=1.8e-12 spinal-cord overlap) depend on external reference gene sets not specified precisely enough to reproduce 1:1.
  • Wet-lab results (qPCR, immunostaining, phagocytosis assays) — not computational.

Compute

All on «infra» «our HPC»-2 via «host» ssh «host»; data on «infra» «path». «host» holds only small results.

Figures / tables: Fig 6AFig 6C
C1
Reported
11119 expressed genes (mean log2 TPM>1.5)
Reproduced
11204
within tolerance
C2
Reported
10463 network genes (CoV>5%)
Reproduced
9380 (closest; CoV def underspecified)
partial
C3
Reported
408 increased / 311 decreased (R47H vs WT basal, Fig 6A)
Reproduced
253 up / 149 down (default); 180/126 (expressed-prefilter)
partial
C4
Reported
184 increased / 83 decreased (IL-4, Fig 6C)
Reproduced
155 up / 53 down (pooled IL-4 vs Ctrl)
partial
C5
Reported
Trem2 decreased in R47H KI, FDR<0.05 (Fig 6A/text)
Reproduced
log2FC=-1.644 padj=2.5e-4 (basal); -1.50 padj=6.3e-9 (pooled)
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 67/100

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: 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 🟡
Concordant (toward reproduced)
Code + data deposited & functional
-2 pts
From: Data & code availability Available & functional
Total score +4

The paper's central claim — Trem2 is suppressed in Trem2-R47H KI microglia — reproduces exactly (log2FC=-1.644, padj=2.5e-4), and the deterministic expressed-gene count matches within 0.76% (11119→11204). The deviations are confined to DEG tallies (C3 408/311→253/149; C4 184/83→155/53) and the network-gene count (C2 10463→9380), all same-direction and ~0.5-0.6x. The cause is on the authors' side: Methods report no software versions, no exact DESeq2 contrast, no salmon flags, and the study's own scripts were not deposited (only the generic CoExpNets tool). No value looks fabricated — all are plausible pipeline outputs — so this is a solid PARTIAL reproduction with explainable, methodology-driven deviations.

🤝
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.

178.4 k
tokens (I/O) · 10.5 M incl. cache
30 min
runtime · 2.29 CPU-h
4.1 GB
peak RAM
3 (1 failed)
HPC jobs
hummel
machine