github_exports_datasets/README.md

UGA/MAF Manuscript Dataset Bundle

This sibling folder stores the readable Hugging Face-ready asset bundle for github_exports.

The primary manifest is dataset_assets_manifest.csv; huggingface_asset_manifest.csv is the compact reader-facing upload manifest. Files are organized by purpose instead of by checksum:

• datasets/: TCGA/MC3, TCGA-BRCA HRD, KUCAB/Mendeley, PCAWG/PanCancer, ClinVar, cBioPortal, Cell Model Passport, CARDIoGRAM, and other curated datasets.
• references/: GRCh37 and GRCh38 reference assets.
• resources/: Bio MAF v4 resources and COSMIC/signature resources.
• results/: large manuscript and result-table artifacts restored into the GitHub checkout.
• caches/: restorable cache/checkpoint assets used by strict validation.
• benchmarks/ and optional_archived/: optional provenance assets.

From the github_exports repository root, validate this folder with:

python scripts/reproduce_manuscript.py --check-datasets --datasets-dir ../github_exports_datasets

Restore active dataset assets into a GitHub checkout and run strict validation with:

python scripts/reproduce_manuscript.py --restore-datasets --datasets-dir ../github_exports_datasets --strict

Current bundle: 4,575 unique SHA-256-validated payloads, 174 active restore destinations, and optional external-project datasets retained for provenance. large_assets_manifest.csv is retained only as a compact large-file planning view.