# GSE293514 source-recovery and validation request

- Recovery ID: `RECOVERY:GSE293514:IDENTITY_ARTIFACTS:1.0`
- Status: **AUTHOR_CONTACT_OR_ARCHIVE_RECOVERY_REQUIRED**
- Official source: [https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE293514](https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE293514)
- Checked: 2026-08-09

## Why recovery is required

The current deposited metadata has 12 columns and no guide identity fields. The matrix exposes 62,711 feature columns while the frozen Ensembl feature inventory has 62,710 rows. Until the author-ordered feature table and guide-to-cell identity are restored, expression-level transfer is blocked because a one-column or barcode misassignment cannot be ruled out.

## Requested artifacts

1. **all_genes.csv** — Exact feature row/column order used with count_matrix.mtx. Required fields: `feature_index`, `ensembl_id`, `gene_symbol`, `reference_build`, `ensembl_release`. Acceptance: Feature count and ordered IDs reconcile exactly with the matrix; any one-column offset fails the gate.
2. **cell_metadata_2000_new.csv** — Exact author-processed cell, sample and guide identity object. Required fields: `barcode`, `sample_2`, `guide_infect`, `guide3`, `guide4`, `guide_type`, `AKAcluster`, `sig_group`, `sig_group2`, `AKAcluster2`, `human_disease`. Acceptance: All matrix barcodes are conserved and guide fields are populated or explicitly marked non-targeting/unassigned.
3. **final_guide_reference.csv** — Guide sequence, target identity and control definitions. Required fields: `guide_id`, `guide_sequence`, `target_gene`, `guide_type`, `control_class`, `positive_control_rule`. Acceptance: MYOD/MYMK positive-control rules, non-targeting controls and duplicate-guide rules are explicit.
4. **cropseq_run_barcode_mapping.csv** — Mapping between eight guide-only runs, transcriptome sublibraries, Parse barcode keys and final bc_wells IDs. Required fields: `run_id`, `library_id`, `parse_barcode_key`, `bc_wells`, `cell_barcode`, `guide_barcode`. Acceptance: Each retained cell has one auditable path from guide run to transcriptome cell identity.
5. **author_processed_fig6_output.h5ad** — Author-processed AnnData or exact Fig. 6 processing output, if available. Required fields: `obs`, `var`, `X`, `uns.processing_provenance`. Acceptance: Reproduces published Fig. 6 cell/cluster counts within the author-defined filtering contract.

## Minimum provenance

- SHA-256 for every returned file
- Parse pipeline version, Scanpy version and reference build
- Confirmation that files correspond to the GEO GSE293514 archive version dated 2026-01-28
- Confirmation that no patient-identifying information is included

## Validation sequence

1. Hash-freeze every returned file and record the source/archive date.
2. Reconcile exact feature count, order, IDs and matrix dimensions.
3. Reconcile every cell barcode and sample assignment.
4. Validate guide-to-cell mapping, single-guide assignment rates and unassigned fraction.
5. Verify non-targeting and MYOD/MYMK positive-control representation.
6. Reproduce author-defined filtering and Fig. 6 cell/cluster counts.
7. Only then assess whether any object can support healthy-myoblast context transfer; no DMD claim is implied.

## Claim boundary

GSE293514 remains a healthy-myoblast screen/context object until feature identity, guide identity and processing provenance are restored. Public accession alone does not authorize expression-level transfer.

Direct DMD candidate perturbation truth remains **0/21**.