Reported software
Cell Ranger 6.0.0
human GRCh38/hg38 pre-mRNA genome
GSE288958 · reference audit 2026-09-09
Reference context is ready; its remaining unknowns stay visible.
This audit reconciles the GEO matrices, opened Seurat object, sample-level QC, donor-wise Scrublet sensitivity and an independent marker-label audit. It closes the context-release checks without claiming ambient-RNA removal, author-exact reprocessing or a DMD perturbation result.
11biopsy/sample units
59,237broad-gate cells
1,579Scrublet candidates
75.89%resolved-label concordance
310kept unresolved
Decision ledger
Each pass states what it closes and what it cannot prove.
The context route has no remaining blocking input. Ambient RNA and author-exact execution stay as enhancement gates; disease-effect and model-label uses remain locked.
| Gate | State | Blocks context | Evidence | Remaining boundary |
|---|---|---|---|---|
official_source_matrix_integrity | PASS | Yes | GEO archive 32 members plus the separate BMD_1 feature table reconcile 11 matrices and 61,189 barcodes | Released matrices contain selected barcodes; they are not unfiltered droplet matrices |
official_processing_metadata | PASS PARTIAL OFFICIAL | Yes | GEO SOFT reports Cell Ranger 6.0.0 and a GRCh38/hg38 pre-mRNA reference for all 11 samples | Exact author command, read structure, whitelist, reference checksum, sample sheet and FASTQ aggregation manifest remain unavailable |
object_and_identifier_reconciliation | PASS | Yes | Seurat opened read-only; 59,222 cells × 29,242 features; cell, sample and feature identities reconcile | Preserve the source object hash and the 1,967-cell raw-to-object denominator gap |
sample_unit_contract | PASS SAMPLE BIOPSY LEVEL | Yes | 11 GSM accessions map one-to-one to 11 BioSample accessions and named biopsy/sample units | Subject-level identifiers are not published; use biopsy/sample as the independent unit and keep site, age and disease explicit |
read_only_matrix_qc | PASS WITH SENSITIVITY | Yes | 59,237 of 61,189 matrix cells pass the frozen broad gate; Normal_2 is retained and predeclared for exclusion sensitivity | The broad gate is an independent audit and does not claim author-exact filtering |
independent_doublet_audit | PASS INDEPENDENT SENSITIVITY | Yes | Donor-wise Scanpy Scrublet maps 59,222/59,222 object cells and flags 1,579 (2.67%); isolated rerun is byte-identical | Algorithmic candidates remain a sensitivity layer; the baseline object is not overwritten and calls are not author labels |
independent_annotation_audit | PASS INDEPENDENT AUDIT BOUNDED | Yes | Independent fixed-marker scoring covers 59,222/59,222 object cells; 58,912 resolve and 44,706 agree with harmonized inherited labels (75.89%) | Keep 310 unresolved cells unresolved; B/plasma lacks an independent marker panel, and disagreements remain visible rather than being overwritten |
ambient_rna | OPEN UNFILTERED DROPLETS UNAVAILABLE | No | A low-count expression profile is available only as a proxy; released matrices do not identify empty droplets | Requires raw_feature_bc_matrix/unfiltered droplets or author contamination estimates for emptyDrops/SoupX closure |
author_exact_reprocessing | OPEN OPTIONAL AUTHOR CONFIRMATION | No | The public processing version and genome family are known, while all six exact execution artifacts remain absent after a provenance-aware local audit | Required only before claiming author-exact Cell Ranger reproduction; it is not required for the bounded read-only context release |
disease_effect_and_model_labels | LOCKED NOT AUTHORIZED | No | No disease-effect p-value, causal contrast, candidate-conditioned DMD outcome or training label is created by this audit | A separate donor/sample-aware statistical contract and outcome-specific validation are required |
Official processing record
Known public recipe and unavailable exact parameters are separate fields.
Released source
11 samples · 61,189 barcodes
The 32-member GEO tar plus the separate BMD_1 feature table reconcile every matrix dimension.
Exact reproduction
Author confirmation still required
Command, read structure, whitelist, reference checksum, sample sheet and FASTQ aggregation manifest are not available.
Unit rule. The public record supports 11 GSM/BioSample biopsy units. Subject-level identity and batch are unavailable, so the independent unit remains biopsy/sample and cells remain nested observations.
Inherited labels × independent marker audit
Agreement and disagreement are both release objects.
The inherited object labels are preserved. Analyst-generated fixed-marker labels provide an independent audit, retain 310 unresolved cells and never overwrite the source object.
| Inherited object label | Harmonized label | Cells | Resolved | Unresolved | Concordant | Concordance | Most common independent label |
|---|---|---|---|---|---|---|---|
B_plasma | b_plasma | 577 | 542 | 35 | 0 | 0.00% | myonuclei · 435 |
endothelial | endothelial | 5,016 | 4,994 | 22 | 3,336 | 66.80% | endothelial · 3,336 |
FAP_fibroblast | fibroblast | 9,294 | 9,282 | 12 | 6,841 | 73.70% | fibroblast · 6,841 |
macrophage_monocyte | macrophage | 4,332 | 4,257 | 75 | 896 | 21.05% | myonuclei · 2,803 |
myonuclei_contractile | myonuclei | 30,395 | 30,393 | 2 | 30,391 | 99.99% | myonuclei · 30,391 |
pericyte_smooth | pericyte | 1,822 | 1,814 | 8 | 844 | 46.53% | pericyte · 844 |
satellite_myogenic | satellite | 5,704 | 5,647 | 57 | 2,194 | 38.85% | myonuclei · 2,506 |
T_NK | t_nk | 2,082 | 1,983 | 99 | 204 | 10.29% | myonuclei · 1,592 |
Known panel gap. No independent B/plasma marker panel is present in the frozen audit method; B_plasma concordance therefore cannot be established by this audit. The full confusion table is downloadable so weak immune, satellite, pericyte and endothelial concordance cannot be hidden by the overall rate.
Independent doublet sensitivity
Scrublet was run per sample and reproduced byte-identically.
59,222 object cells map exactly; 1,579 (2.67%) are algorithmic candidates. They remain a sensitivity layer.
| Sample | Object cells | Candidates | Fraction | Median score | Automatic threshold |
|---|---|---|---|---|---|
| Normal_1 | 9,165 | 509 | 5.55% | 0.0417 | 0.1899 |
| Normal_2 | 3,020 | 40 | 1.32% | 0.0642 | 0.3141 |
| Normal_3 | 6,642 | 159 | 2.39% | 0.0230 | 0.2519 |
| Normal_4 | 3,774 | 82 | 2.17% | 0.0413 | 0.2642 |
| Normal_5 | 4,621 | 72 | 1.56% | 0.0309 | 0.2563 |
| BMD_1 | 5,518 | 121 | 2.19% | 0.0205 | 0.2295 |
| BMD_2 | 8,613 | 319 | 3.70% | 0.0660 | 0.2543 |
| BMD_3 | 5,323 | 111 | 2.09% | 0.0656 | 0.2871 |
| DMD_1 | 3,368 | 29 | 0.86% | 0.0262 | 0.2755 |
| DMD_2 | 3,062 | 7 | 0.23% | 0.0234 | 0.4590 |
| DMD_3 | 6,116 | 130 | 2.13% | 0.0200 | 0.2379 |
Frozen method: Scanpy 1.12.1 · expected rate 0.05 · simulation ratio 2.0 · 30 PCs · seed 20260720 · isolated rerun IDENTICAL.
Read-only sample QC
Normal_2 remains visible and predeclared for sensitivity.
The broad gate is an independent matrix audit. It does not recreate the authors’ filtering choices.
| Sample | Matrix cells | Retained | Retained | Median counts | Median features | Policy |
|---|---|---|---|---|---|---|
| Normal_1GSM8779827 · control | 9,170 | 9,166 | 99.96% | 4,140 | 1,851 | Retained |
| Normal_2GSM8779828 · control | 4,953 | 3,029 | 61.15% | 582 | 403 | Predeclared sensitivity |
| Normal_3GSM8779829 · control | 6,644 | 6,644 | 100.00% | 2,528 | 1,145 | Retained |
| Normal_4GSM8779830 · control | 3,775 | 3,774 | 99.97% | 5,197 | 2,164 | Retained |
| Normal_5GSM8779831 · control | 4,623 | 4,621 | 99.96% | 4,463 | 1,912 | Retained |
| BMD_1GSM8779832 · BMD | 5,521 | 5,521 | 100.00% | 1,332 | 859 | Retained |
| BMD_2GSM8779833 · BMD | 8,615 | 8,613 | 99.98% | 2,814 | 1,311 | Retained |
| BMD_3GSM8779834 · BMD | 5,341 | 5,323 | 99.66% | 3,233 | 1,629 | Retained |
| DMD_1GSM8779835 · DMD | 3,368 | 3,368 | 100.00% | 1,874.5 | 1,225 | Retained |
| DMD_2GSM8779836 · DMD | 3,063 | 3,062 | 99.97% | 4,742 | 2,235 | Retained |
| DMD_3GSM8779837 · DMD | 6,116 | 6,116 | 100.00% | 2,965.5 | 1,625 | Retained |