Single-cell protein activity analysis reveals aberrant myogenesis and IGF2-PI3K pathway dependencies in MYOD1-mutant rhabdomyosarcoma.
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 are derivable from the shared data
- ✓Any deviation was negligible
- 🟡Reported values were only indirectly comparable
- 🟡A deviation arose in the data or preprocessing
- 🟡A deviation was attributed to the published material
- 🟡The central claim did not (fully) hold under reproduction
- 🟡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 for the QC front-end -> 1:1 within ~1.7%. Reproduced the deterministic QC step of the authors' own Seurat pipeline (Read10X + CreateSeuratObject(min.cells=3, min.features=200)) on the paper's OWN public data (GEO GSE288065). Total post-QC nuclei over the 6 patient samples = 189,250 vs the paper's implied total ~192,447 (177,051 malignant / 0.92), i.e. within 1.66%; equivalently 0.92x189,250 = 174,110 vs 177,051 reported malignant (1.66%). Per-sample counts recorded. No fabrication signal: the deposited raw matrices support the reported cohort size under the stated QC. NOT attempted (out of scope / 80-20 skip): the deep protein-activity results -- ARACNe-AP networks, VIPER master regulators (IGF2/PRKG1/ANKRD1), OncoTarget/IGF2-PI3K, SingleR malignant fraction (92%), and Seurat-v4 SCT integration/subclustering (20 subclusters; 3 states 68265/82053/26292) -- because the repo does not ship functions/*.R helpers, regulator ensembl lists, oncotarget.csv, mart.obj.rds, or the Tabula Sapiens muscle reference, and these are the expensive stochastic last ~20%. Data-accession correction: the room brief auto-enriched GSE195709 (a DIFFERENT paper, PMID 35982179); the correct dataset is GSE288065 (repo README + GEO pubmed_id). All wet-lab/genomic-cohort/in-vivo results are non-pipeline and out of scope.
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 84assessed: 2026-06-14 ⛓ 26210175fc33
✎ 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
- 2026-09-19
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: sonnetMYOD1 L122R-mutant spindle cell/sclerosing rhabdomyosarcoma harbors distinct, disrupted myogenic differentiation programs driven by mutant MYOD1, and single-cell regulatory network analysis can reveal actionable master regulator (MR) dependencies—particularly IGF2-IGF1R-PI3K/AKT/mTOR signaling—that can be therapeutically targeted.
- ★ MYOD1 L122R-mutant SRMS tumors contain three coexisting, conserved cell states (progenitor, transition, differentiated) reflecting aberrant myogenic differentiation finding
- ★ A paracrine IGF2-IGF1R-PI3K/AKT/mTOR signaling axis from progenitor to transition/differentiated cells sustains PI3K/AKT/mTOR pathway activity mechanism
- ★ IGF1R and PI3K/AKT/mTOR pathway inhibition shows selective sensitivity in ex vivo PDX-derived organoid drug screens and improves disease control in a PDX model in vivo finding
- ★ Single-cell metaVIPER/ARACNe protein activity inference robustly identifies oncogenic master regulators from sparse snRNA-seq data in an ultrarare tumor method
- ★ MYOD1 displays a largely distinct regulon/target profile relative to other myogenic factors, consistent with altered regulatory function due to the L122R mutation mechanism
- Oncogenic MRs identified in snRNA-seq are recapitulated in an independent cohort of 24 bulk RNA-seq profiles finding
- Targeted DNA sequencing of 20 profiles reveals recurrent IGF2/PI3K/AKT pathway alterations reinforcing shared transcriptional vulnerabilities finding
- ★ Combination of a PI3K/mTOR inhibitor with chemotherapy produces objective tumor regression in a PDX model finding
| Assay | System | Perturbation | Readout | Platform |
|---|---|---|---|---|
| single-nucleus RNA sequencing (snRNA-seq) | 6 patient-derived, chemotherapy-exposed, snap-frozen MYOD1 L122R-mutant SRMS tumors | none | gene expression profiles, malignant nuclei identification, cell state composition | — |
| InferCNV copy number inference | tumor nuclei vs. sample-matched immune cells (singleR-annotated) | none | large-scale somatic copy number alterations to distinguish malignant vs nonmalignant nuclei | InferCNV |
| ARACNe-AP gene regulatory network reconstruction | tumor cell gene expression data from six MYOD1 L122R-mutant SRMS | none | sample-specific regulatory interactions (regulons) between transcriptional regulators and target genes | ARACNe-AP |
| metaVIPER protein activity inference | single tumor nuclei from snRNA-seq | none | inferred activity of 2336 regulatory proteins per cell | metaVIPER (VIPER algorithm) |
| immunohistochemistry | paired pretreatment biopsies and posttreatment resections from patient tumors | neoadjuvant chemotherapy (VAC) | MYOD1 and myogenin protein expression | — |
| targeted tumor-normal DNA sequencing (MSK-IMPACT) | patient tumor/normal DNA (4 of 6 snRNA-seq cases initially; 20 total DNA profiles) | none | MYOD1 variant allele frequency, co-occurring genomic alterations (e.g., PIK3CA, PI3K/AKT/mTOR pathway) | MSK-IMPACT 341-505 gene panel |
| bulk RNA sequencing | independent cohort of 24 MYOD1 L122R-mutant SRMS tumor samples | none | validation of oncogenic master regulator activity | — |
| ex vivo drug screen and in vivo xenograft treatment | chemotherapy-resistant patient-derived xenograft (PDX) cell cultures/organoids (PDXOs) and PDX mouse model | IGF1R and PI3K/AKT/mTOR pathway inhibitors, alone or combined with chemotherapy | drug sensitivity/viability; in vivo tumor growth, disease control, tumor regression | — |
- – Integrated snRNA-seq dataset contained 177,051 high-quality malignant nuclei, the largest single-cell dataset generated for this rare sarcoma 177,051 nuclei
- – Three consensus tumor cell states identified: progenitor (68,265 nuclei), transition (82,053 nuclei), differentiated (26,292 nuclei), consistently observed across patients 68,265 / 82,053 / 26,292 nuclei
- ▲ MYOD1 variant allele frequency exceeded 60% across MSK-IMPACT-profiled samples >60% VAF
- ▲ IGF2-IGF1R-PI3K paracrine signaling from progenitor to transition/differentiated states sustains PI3K/AKT activity in those states
- – MYOD1 diffusely and strongly positive by IHC in all samples, while myogenin was rare to negative
- ▼ PI3K/mTOR inhibitor combined with chemotherapy produced objective tumor regression in vivo; single agent achieved meaningful disease control
- ▲ 30 to 40% of MYOD1 L122R-mutant SRMS tumors harbor recurrent comutations, most commonly PIK3CA hotspot mutations, in the PI3K/AKT/mTOR pathway 30-40%
- – Malignant nuclei constituted 92% of the entire dataset, with nonmalignant stromal/immune cells representing 6 to 27% across samples 92% malignant; 6-27% nonmalignant
- count 177,051 malignant nuclei (integrated snRNA-seq dataset across six tumors)
- count 20 subclusters (two to five per tumor) (unsupervised Louvain clustering of metaVIPER protein activity across six tumors)
- count progenitor 68,265; transition 82,053; differentiated 26,292 nuclei (three consensus cell states)
- other malignant nuclei 92% of dataset; nonmalignant 6-27% across samples (tumor vs nontumor nuclei composition)
- count 2336 regulatory proteins (metaVIPER-inferred protein activity per cell)
- other MYOD1 VAF >60% (MSK-IMPACT-profiled samples)
- count 24 bulk RNA profiles (independent validation cohort for oncogenic MRs)
- count 20 DNA profiles; PIK3CA/PI3K-AKT-mTOR comutations in 30-40% of tumors (targeted DNA sequencing cohort)
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 applied single-nucleus RNA sequencing (snRNA-seq) to six patient-derived MYOD1 L122R-mutant rhabdomyosarcoma tumors (177,051 malignant nuclei), constructing sample-specific gene regulatory networks via ARACNe-AP and inferring protein activity at single-cell resolution using metaVIPER. Unsupervised resolution-optimized Louvain clustering followed by pairwise viperSimilarity (analytic rank-based enrichment analysis, aREA) defined three consensus tumor cell states. Findings were extended to 24 bulk RNA and 20 DNA profiles, and functionally validated in patient-derived xenograft (PDX) cell cultures, organoids, and an in vivo PDX model.
| Test | Applied to | n | Assumptions |
|---|---|---|---|
| Analytic rank-based enrichment analysis (aREA) via viperSimilarity — normalized enrichment score (NES) of shared master regulator overlap | Pairwise comparison of 20 subcluster mean protein activity signatures to define three consensus cell states | 20 subclusters derived from 177,051 malignant nuclei across 6 tumors | not stated |
| Resolution-optimized Louvain clustering on VIPER-inferred protein activity | Identification of 20 patient-specific tumor subclusters | 177,051 malignant nuclei across 6 tumors | not stated |
| InferCNV (single-nucleus copy number inference) | Distinguishing malignant from nonmalignant nuclei per sample | All nuclei per sample, referenced against sample-matched immune cells annotated by singleR | not stated |
| CytoTRACE differentiation scoring | Quantifying differentiation potential across tumor cell states | 177,051 malignant nuclei | not stated |
| singleR automated cell-type annotation | Annotating nonmalignant (lymphoid, myeloid, stromal) versus malignant nuclei | — | not stated |
| Seurat cell-cycle scoring (S and G2/M phase scores) | Estimating proliferative fraction across tumor cell states | 177,051 malignant nuclei | not stated |
-
Cell states were defined by hierarchical clustering of viperSimilarity NES matrices followed by refinement using differentiation markers, with the authors explicitly noting states 'lay along a continuum rather than forming discrete clusters'↳ Could also: Trajectory inference methods such as Monocle 3, PAGA, or diffusion pseudotime could also model continuous differentiation gradients, providing pseudotime ordering and branch probabilities along the progenitor-to-differentiated axis — When the underlying biology is a continuum, trajectory-based approaches make that structure explicit and allow quantification of each cell's position along the differentiation axis, complementing the discrete state assignments used here
-
Protein/transcription-factor activity was inferred de novo from ARACNe-reconstructed, sample-specific regulons via metaVIPER↳ Could also: Curated regulon resources such as DoRothEA/decoupleR or pySCENIC (which builds co-expression regulons from the data itself) could also estimate transcription factor activity and serve as an orthogonal benchmark — Curated regulon databases provide a data-independent reference that can cross-validate de novo network findings, which is particularly informative in a six-sample rare-tumor setting where the ARACNe network may be sensitive to sample composition
-
Malignant cell identification relied on InferCNV copy number inference referenced against sample-matched immune cells↳ Could also: CopyKAT, SCEVAN, or Numbat (which additionally leverages allele-specific phasing) could also infer copy number states from snRNA-seq to classify malignant nuclei, with orthogonal confirmation possible from matched bulk DNA data — Different CNV-inference tools carry different sensitivity/specificity tradeoffs in sparse nuclear data; applying a second tool or comparing to matched bulk CNV profiles strengthens confidence in malignant cell assignments
-
The six-tumor snRNA-seq cohort size is stated without a formal sample size justification or power analysis↳ Could also: Post-hoc saturation analysis (rarefaction of nuclei or tumors) or leave-one-out stability testing of the three cell states could also quantify how robust the identified states are to the small patient n — In rare-disease studies where prospective power calculations are infeasible, saturation and jackknife analyses provide an empirical estimate of whether the main findings are stable across the cohort
-
2336 regulatory proteins were assessed per cell for master regulator nomination, with no explicit multiple-testing correction described in the available text↳ Could also: Benjamini-Hochberg FDR correction applied across the family of regulators tested per cell-state comparison, or permutation-based FDR as commonly used in GSEA-type enrichment frameworks, could also control the expected false-discovery rate — Explicit FDR control quantifies how many nominated master regulators are expected to be false positives; when thousands of proteins are tested, reporting an adjusted threshold helps readers calibrate confidence in the candidate list
-
Ex vivo drug screen sensitivity and in vivo PDX treatment outcomes were used to validate therapeutic targeting; effect magnitude is described qualitatively in the abstract ('significantly improved disease control', 'objective tumor regression')↳ Could also: Quantitative effect sizes such as area under the dose-response curve (AUC), tumor growth inhibition percentage, or hazard ratios with 95% confidence intervals alongside p-values from mixed-effects or repeated-measures models could also summarize drug response magnitude — Reporting both a p-value and a standardized effect size with uncertainty interval allows readers to assess translational relevance separately from statistical significance, which is especially informative for small PDX cohorts where effect size rather than p-value drives clinical interpretation
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-41758938
Paper: Dermawan JK et al. Single-cell protein activity analysis reveals aberrant myogenesis and IGF2-PI3K pathway dependencies in MYOD1-mutant rhabdomyosarcoma. Sci Adv 2026. PMID 41758938 · PMCID PMC12947870 · DOI 10.1126/sciadv.aea6453. Code: https://github.com/dermawaj/MYOD1snRNA @ 75db9f8 (MIT, pushed 2026-02-13).
Data accession correction (IMPORTANT)
- Repo README and GEO both give the dataset as GSE288065 (snRNA-seq, 6 MYOD1-mutant
spRMS patients, 7 GSMs GSM8758451–8758457, public since 2026-01-22,
pubmed_id=41758938). - The room brief's auto-enrichment recorded GSE195709, which is a different paper (Wei/Langenau, "Stem cell and developmental hierarchies in RMS", PMID 35982179, 2022). → We use GSE288065 (the correct, paper-linked dataset).
Pipeline-derived results (candidate in-scope)
The repo ships R/shell scripts implementing a standard Seurat→ARACNe-AP→VIPER stack:
| step | script | pipeline | reproducible? |
|---|---|---|---|
| QC + SCTransform + Seurat-v4 integration | seurat_RMS_pipeline.R | Seurat | front-end YES (deterministic QC); integration/clustering heavy + stochastic |
| cell typing (malignant vs TME) | seurat_RMS_pipeline.R | SingleR + BlueprintEncode (celldex) | partial (ref downloadable; label set stochastic) |
| ARACNe networks (per sample, TF/coTF/sig/surface × 100 bootstraps) | ARACNe.sh | ARACNe-AP (java) | heavy; needs ensembl regulator lists (tfs/cotfs/sig/surface-ensembl.txt) NOT shipped |
| VIPER protein activity + master regulators | tumorigenicMR.R | viper | needs ARACNe nets + Tabula Sapiens muscle ref + mart.obj.rds NOT shipped |
| OncoTarget (IGF2/PI3K druggable MRs) | tumorigenicMR.R | viper + oncotarget.csv | needs oncotarget.csv NOT shipped |
Missing-from-repo dependencies (block full reproduction)
seurat_RMS_pipeline.R and tumorigenicMR.R source("functions/{process-utils, cluster-functions,viper-utils,misc}.R") — only misc.R (root) is in the repo; the
functions/ dir is absent. Also absent: regulator lists (*-ensembl.txt),
oncotarget.csv, mart.obj.rds, tabulaSapiens_muscle.rds, MYOD1_COUNT.txt
(bulk), and the subcluster/group CSVs. → The deep VIPER/MR/OncoTarget results are
not runnable as shipped (docs/code-insufficient for the last ~20%).
IN SCOPE (attempted — clearly specified, deterministic, low-hanging)
- Post-QC nuclei count from the paper's own GEO matrices using the authors'
exact QC:
Read10X+CreateSeuratObject(min.cells=3, min.features=200). Methods: "Nuclei expressing fewer than 200 genes and genes detected in fewer than three nuclei were filtered out." Reported anchor: 177,051 malignant nuclei (= 92% of dataset → implied total ≈ 192,447). Pipeline: Seurat. - Sample count = 6 (trivial structural check).
OUT OF SCOPE (not attempted, with reason)
- ARACNe-AP network reconstruction & VIPER protein activity / master regulators
(IGF2, PRKG1, ANKRD1), OncoTarget/IGF2-PI3K druggable-MR heatmaps → 80/20 skip:
required regulator lists + reference objects +
functions/code are not shipped; ~2.4k bootstrap ARACNe runs × VIPER is the expensive last 20%. - Seurat-v4 SCT integration + Louvain subclustering (20 subclusters; 3 states with 68,265 / 82,053 / 26,292 nuclei) → stochastic, depends on unshipped helper code.
- SingleR malignant-fraction (92%) → needs celldex ref + helper code; deferred.
- All wet-lab / genomic-cohort / in-vivo PDX results (Figs 7–8; PIK3CA 20%, PTEN 10%) → non-pipeline, out of scope by definition.
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.
We reproduced only the deterministic QC front-end of the authors' Seurat pipeline on their own public GEO data (GSE288065): 189,250 nuclei vs the implied 192,447, and 174,110 vs 177,051 malignant — both within 1.66%, with no fabrication signal and the residual explained by mito/doublet QC we did not apply (our-method side, input-level). The paper's central thesis — IGF2-PI3K pathway dependency and aberrant myogenesis derived from ARACNe/VIPER protein-activity master regulators — was not attempted because required helper code and reference objects are unshipped, so the core claim is untested rather than refuted. Overall a solid but partial reproduction: cohort counts confirmed, deep biology unverified.
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 [email protected].
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.