# Data attribution and changes

The current canonical schema 3 release uses the official [RPPA500 patient download](https://tcpa.drbioright.org/rppa500/download.html), the TCPA patient/cell-line antibody metadata, and the [NCI GDC public case catalogue](https://api.gdc.cancer.gov/cases) for participant-to-cancer-project annotations missing from TCPA's expression metadata. The public TCGA catalogue was fetched without submitting local participant identifiers; joins occur locally by exact source identifiers. URLs, response hashes and retrieval times are preserved in `source-cache/acquisition.json` and the source lock. [The RPPA500 study](https://pmc.ncbi.nlm.nih.gov/articles/PMC11999798/) describes the resource. These additions are distinct from the historical PanCan11 cohort and are included in this canonical release.

The data artifacts in this repository derive from the local TCPA downloads named and SHA-256 pinned in `../build_manifest.json`. Per-dataset release, normalization and source metadata are retained in `../manifest.parquet` and sample `source_meta_*` columns. Missing release information is explicitly unknown. The release includes all 64 catalogued CPPA sets, including set120; preserve each set's source release and normalization fields.

Published drug-analysis results come from the public [cell-line](https://tcpa.drbioright.org/rppa500mclp/#/) and [CPPA](https://tcpa.drbioright.org/cppa/#/) portal databases, retrieved on 2026-10-08. Each document's URL, revision, hash and retrieval time is preserved in `source-cache/portal-analyses/manifest.json` and inside the corresponding store. Values are unchanged; the sensitivity datasets they summarize (CTRPv2, GDSC, MDACC) are not redistributed here.

**CONTRIB-TCPA00000009 / TCPA00000009:** © Satoshi Nishizuka and Hidewaki Nakagawa. Licensed under [Creative Commons Attribution 4.0 International](https://creativecommons.org/licenses/by/4.0/). The supplied licence is preserved verbatim in `CONTRIB-TCPA00000009-LICENSE.txt` alongside this notice, and its raw input hash is recorded in the build manifest.

Changes to that contributed dataset: source metadata and dataset-qualified sample/assay keys are added; source numeric measurements are preserved exactly. It remains in the source inventory and contributed matrix; the current patient/contributed notebook defaults to RPPA500-TCGA and permits selection of all 13 contributed datasets, including historical PanCan11. No endorsement by the original creators is implied.

Other source data retains its source-specific metadata; this notice does not relicense all TCPA data under the licence of TCPA00000009. Producer code, raw inputs and historical validation evidence remain in the source repository. This bundle preserves the final canonical data; it does not rerun acquisition or scientific validation.

This publication notice corrects stale scope/default descriptions in the source notice; data bytes and the supplied dataset09 licence are unchanged. Paths under source-cache refer to producer-side evidence, not files required by the notebooks. Snapshot: tcpa-data-2026-10-08-a19c482.

Deployment derivative tcpa-data-2026-10-08-a19c482-slash-v1 preserves scientific data; see RELEASE.json and LAYOUT_CHANGES.json for the physical key-layout change.