Endothelial Adgrl2 Expression and Alternative Splicing Controls the Cerebrovasculature.
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
- ✓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
- 🟡Reported values were only indirectly comparable
- 🟡Overall, the reproduction showed a material discrepancy
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 to reproduce the in-scope pipeline results 1:1 from public data. The paper's only code link (rvalieris/parallel-fastq-dump) is a generic third-party FASTQ downloader, NOT the authors' analysis code (none released); per the P16 rule we re-derived the pipeline-derived results from the paper's own public input data (Allen Brain Map SMART-seq = GEO GSE185862) using the Methods-specified normalization (CPM -> log2(CPM+1)) and Tau specificity. RESULT: Table 1 reproduced EXACTLY (all 42 cell subclasses and all per-subclass counts, total 73347, match the metadata.csv 1:1 -> no fabrication signal). Fig 1 central claim reproduced directionally: scoped to non-neuronal populations (the paper's exact wording), Adgrl2 is the #1 endothelial-enriched latrophilin, with Adgrl1 near-absent in endothelium and Adgrl3 broadly expressed -- matching the paper; graded 'partial' only because Fig 1 prints no numeric values so an exact numeric 1:1 is impossible. NOT ATTEMPTED (hard 20%, documented in scope.md): the alternative-splicing EIP analysis (Fig 3-4) requires multi-TB raw FASTQ from SRA, STAR realignment to GRCm39, and an UNRELEASED custom Python splice-junction counter; and all wet-lab results (smFISH, IHC, BBB assay, vascular morphology) are non-computational. Compute on «our HPC» SLURM (account kubisch_std); data kept on «infra», only small results on «host».
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 75assessed: 2026-06-14 ⛓ 3ee9e356d055
✎ 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
- 2026-06-14
- Rubric version
- v1.0
- Assessed by
-
🤖 AI curator · claude (ai-curator room) · v1.0 · run #1 2026-06-15no 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: opusBecause Adgrl2 is expressed in both neurons and brain endothelial cells, the paper tests how a single latrophilin gene can serve distinct functions in neural circuit assembly versus cerebrovascular homeostasis, hypothesizing that cell type-specific alternative splicing produces distinct Adgrl2 isoforms underlying these separable functions.
- ★ Endothelial cell-specific Adgrl2 deletion impairs cerebrovascular integrity (blood–brain barrier). finding
- ★ Adgrl2 mRNA undergoes robust cell type-specific alternative splicing, producing distinct isoforms in neurons versus endothelial cells. finding
- ★ Forcing expression of the neuronal Adgrl2 isoform in endothelial cells induces ectopic glutamatergic synaptic contacts onto endothelial cells and enhances BBB integrity. finding
- ★ Overly restrictive endothelial Adgrl2 function dysregulates blood-to-CSF homeostasis, enlarges brain ventricles, and increases hydrocephalus risk. finding
- ★ Alternative splicing is a cell type-specific mechanism providing isoform-specific Adgrl2 for discrete functions in neural circuit assembly and cerebrovascular homeostasis. mechanism
- Among latrophilins (Adgrl1-3), the greatest splicing/expression contrast is between neurons and non-neuronal cells, with Adgrl2 most polarizing. finding
- Exon Inclusion Proportion (EIP) and Tau specificity metrics applied to scRNAseq splice-junction reads to quantify cell type-specific splicing. method
- Adgrl2 (Lphn2) expression in non-neuronal brain cells is restricted to endothelial cells. finding
| Assay | System | Perturbation | Readout | Platform |
|---|---|---|---|---|
| single-molecule RNA fluorescent in situ hybridization (smFISH) | P30 wild-type mouse brain cryosections | none | Adgrl2/Lphn2 puncta per cell colocalized with cell-type markers (CD31, Pdgfrβ, Aldh1l1) | ACDBio RNAscope probes, Akoya Opal fluorophores, Zeiss 880 confocal with Airyscan |
| single-cell RNA sequencing analysis (alternative splicing / gene expression) | mouse cortex/hippocampus and other published datasets (Allen Brain smartSeqV4; embryonic neurons; riboTRAP; brain/lung endothelial cells) | none | gene CPM and Exon Inclusion Proportion (EIP) from splice junction reads, Tau specificity | STAR aligner, Trimmomatic, GRCm39; SMART-Seq v4 |
| immunohistochemistry / immunofluorescence with 3D confocal analysis | mouse brain horizontal vibratome sections (Tie2-Cre; Adgrl2 fl/KI; Ai14) | endothelial Adgrl2 KO and endothelial expression of neuronal isoform (Tie2-Cre driven) | endothelial (CD31) surface coverage by pericytes (Cspg4) and astrocytes (Aqp4); vGlut1/vGlut2/vGat presynaptic puncta on vessels | Zeiss 880 confocal, Imaris 9.6 |
| blood–brain barrier integrity assay (sodium fluorescein extravasation) | mouse brain and liver tissue | endothelial Adgrl2 deletion / isoform switch | tissue sodium fluorescein concentration | SpectraMax iD5 plate reader (ex 490 nm / em 530 nm) |
| vascular morphology quantitative analysis | mouse brain CD31-labeled vasculature images | Adgrl2 genetic manipulations | vessel morphometrics from segmented CD31 channel | custom Python script (scikit-image, aicsimageio, Shapely) |
- ▼ Endothelial cell-specific Adgrl2 deletion impairs cerebrovascular integrity.
- – Neurons and endothelial cells express distinct Adgrl2 isoforms via cell type-specific alternative splicing.
- ▲ Forced neuronal Adgrl2 isoform in endothelial cells produces ectopic glutamatergic synaptic contacts onto endothelial cells.
- ▲ Endothelial expression of the neuronal isoform enhances blood–brain barrier integrity (opposite of deletion).
- ▲ Restrictive cerebrovascular function enlarges brain ventricles and raises hydrocephalus risk.
- other Adgrl2 puncta quantified if within 5 μm of nucleus (smFISH puncta counting threshold per cell)
- other minimum of 100 cells and detection >1 log2(CPM + 1) (inclusion criteria for analyzing a cell type's alternative splicing)
- other 200 mg sodium fluorescein per kg body weight (25 mg/ml) (dose for BBB integrity assay)
- other rejected if >3 median absolute deviations below median (scRNAseq cell quality filtering)
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 study employs genetic mouse models (endothelial Cre-lox Adgrl2 deletion and neuronal-isoform knockin), re-analysis of published scRNAseq datasets, smFISH, 3D immunohistochemistry quantification, and a sodium fluorescein-based blood-brain barrier permeability assay. scRNAseq data were normalized to counts per million (CPM) and log2(CPM+1)-transformed; tissue-expression specificity was summarized with the Tau metric and alternative splicing with per-cell exon inclusion proportions (EIP). The provided text excerpt does not reach the section describing inferential statistical tests or numerical results, so between-group test choices, dispersion reporting, and p-value handling cannot be characterized from the supplied text.
| Test | Applied to | n | Assumptions |
|---|---|---|---|
| not stated in provided excerpt — the text is truncated before inferential statistical tests for group comparisons are described | all between-group comparisons (KO vs. control; KI vs. control) | — | not stated |
-
Alternative splicing was summarized with a custom per-cell exon inclusion proportion (EIP) derived from splice-junction read counts↳ Could also: Established differential splicing tools such as rMATS, MAJIQ, or MISO could also quantify exon inclusion and provide statistical testing with FDR-corrected p-values and credible intervals across cell-type groups — Dedicated splicing tools incorporate statistical models for read-count variability and overdispersion, yielding calibrated significance estimates that complement a descriptive EIP summary and make comparisons with other published datasets more straightforward
-
Gene expression specificity across cell subclasses was quantified with the Tau metric↳ Could also: Alternative specificity indices such as the Jensen-Shannon divergence-based specificity score or a z-score enrichment approach (analogous to GSEA-style enrichment) could also quantify tissue/cell-type selectivity — Different specificity metrics weight rare high-expressing subclasses differently; comparing conclusions across two metrics (e.g., Tau and JSI) can confirm that specificity calls are not sensitive to the particular formula chosen
-
scRNAseq quality filtering excluded cells more than 3 median absolute deviations (MAD) below the median for mapped-read count and unique-gene count↳ Could also: Standard single-cell QC pipelines (e.g., scater, Seurat) additionally incorporate mitochondrial read fraction as a third filtering dimension alongside depth and gene-count MAD thresholds — High mitochondrial content flags stressed or partially lysed cells that can pass depth-only filters; including it as a QC dimension is a widely adopted complement in single-cell studies
-
smFISH puncta counts per cell were normalized to the image-level mean prior to any between-group comparison↳ Could also: A linear mixed-effects model with image and animal as nested random effects could also account for the hierarchical structure of cells within images within animals — Mean normalization removes systematic image-level offsets but does not propagate the uncertainty arising from the nesting structure; a mixed model yields standard errors that reflect both within-image and between-animal variability, which is particularly relevant when animal n is small
-
Blood-brain barrier permeability was expressed as absolute tissue sodium fluorescein concentration determined from a standard curve↳ Could also: Normalizing brain NaFl to simultaneously collected serum NaFl (a brain-to-serum ratio) is also standard in BBB permeability protocols — Serum normalization corrects for inter-animal differences in injection efficiency, body weight-adjusted dosing accuracy, and renal clearance rate, reducing a source of variability that is independent of true BBB function
-
3D vascular morphology metrics were extracted with a custom Python script using scikit-image, aicsimageio, and Shapely↳ Could also: Established vascular analysis platforms such as AngioTool, the FIJI AnalyzeSkeleton plugin, or MorphoLibJ could also produce comparable morphological endpoints with published validation histories — Custom scripts are fully valid and offer flexibility, but their deposited code and version documentation (the authors provide a Zenodo DOI) serve the same reproducibility goal that established validated tools achieve through community testing; either approach is sound when code is openly shared
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.
What was reproduced
The exact results taken into scope, with each reported value next to the value our attempt produced.
Scope — pmid-41876233
Paper: King A et al., "Endothelial Adgrl2 Expression and Alternative Splicing Controls the Cerebrovasculature." J Neurosci 2026. PMID 41876233 · PMC13108381 · DOI 10.1523/jneurosci.0019-26.2026.
Listed code: https://github.com/rvalieris/parallel-fastq-dump — this is a third-party generic SRA→FASTQ download utility, NOT the authors' analysis code. Per BRIEF rule 2 (P16), applying a third-party tool to the paper's own data is a valid reproduction. The paper ships no analysis code repo of its own; the splice/expression analysis is described in prose ("analyzed using Python 3") with no released scripts.
Listed data: GEO GSE185862 = Allen Institute "A taxonomy of transcriptomic
cell types across the isocortex and hippocampal formation" (Yao et al., Cell 2021).
The paper uses the SMART-seq v4 subset (73,363 cells). Public, fully
downloadable from the Allen Brain Map S3 bucket
(s3://idk-etl-prod-download-bucket/aibs_mouse_ctx-hpf_smart-seq/).
In scope (pipeline-derived, clearly specified — the 80)
| # | Result | Paper location | Pipeline | Data needed |
|---|---|---|---|---|
| C1 | 42 cell subclasses + per-subclass cell counts (total 73,347; ">73,000 cells; 42 subclasses") | Table 1 + Results §1 | Count cells per subclass_label in Allen SMART-seq metadata |
metadata.csv (28 MB) |
| C2 | Adgrl2 is selectively expressed by endothelial cells (cell-type-specific expression, distinct from Adgrl1/Adgrl3) | Fig 1A–C, Results §1 | CPM → log2(CPM+1) per cell from Allen counts (introns+exons), aggregate per subclass; expression-specificity metric Tau | matrix.csv (14 GB) + metadata.csv |
Both are derived directly from the public Allen count matrix + metadata using the exact normalization the Methods specify (per-cell counts / total reads × 1e6 = CPM; then log2(CPM+1); Tau per Yanai et al. 2005). Low-compute, unambiguous, 1:1 checkable by hand.
Out of scope / not attempted (the hard 20%) — documented, not dropped silently
- Alternative-splicing EIP (Exon Inclusion Proportion) analysis (Fig 3–4, splicing program in neurons vs endothelial cells). Requires: downloading raw FASTQ for ~73k SMART-seq cells from SRA (multi-TB), Trimmomatic, STAR realignment to GRCm39/Annotation Release 109, then a custom Python splice- junction counter that is not released. The EIP formula is given but the code is not, and the compute (TB-scale FASTQ + per-cell alignment) is disproportionate. → Skipped per BRIEF rule 3 (80/20). Reason recorded; not a drop of the paper.
- Wet-lab / manual results (smFISH, IHC, blood–brain-barrier assay, vascular morphology, mouse genetics) — not computational, out of scope by rule 2.
- Additional scRNAseq datasets re-analyzed for splicing (Lukacsovich 2019, Furlanis 2019, Vanlandewijck 2018) — same splicing pipeline as above, skipped.
Possible-fabrication checks enabled by this reproduction
- Do the 42 subclass labels + counts in Table 1 match the public Allen metadata exactly? (Any value not derivable from the shipped data is flagged.)
- Is Adgrl2 genuinely endothelial-enriched in the public matrix, as Fig 1 claims?
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.
For the computationally in-scope claims this is a strong reproduction: Table 1 (42 subclasses, 73,347 cells, all per-subclass counts) reproduces exactly 1:1 from the public Allen GSE185862 data, and the Fig 1 endothelial-Adgrl2 specificity claim reproduces directionally (Endo Adgrl2 5.01 log2CPM, #1 of 6 non-neuronal, Tau 0.467; Adgrl1 near-absent, Adgrl3 broad). The only frictions are on our/availability side, not the authors': Fig 1 is graphical-only so an exact numeric match can't be pinned (q2 yellow), and the authors released no analysis code (the linked repo is a generic FASTQ downloader). The splicing (Fig 3-4) and wet-lab results were out of scope and not attempted, so q8 is yellow — solid and well-explained rather than a full end-to-end 1:1 of the whole paper.
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.