This document outlines the master data strategy for the Cytognosis Foundation, aligning with NIH GDS policy and our FAIR data principles.
gs://cytognosis-public-data): De-identified, open-access datasets (e.g., benchmark data, synthetic profiles).gs://cytognosis-internal): Internal development, non-PHI operational data, testing data.gs://cytognosis-data-hub): Processed datasets shared with external collaborators (e.g., Purdue).gs://cytognosis-phi-core): Raw genomic/clinical PHI, L3 controlled data.gs://cytognosis-phi-collab-nih): NIH Data Use Certification (DUC) governed collaboration data.Cytognosis enforces FAIR (Findable, Accessible, Interoperable, Reusable) principles via the following stack:
1. FAIRDOM-SEEK (hub.cytognosis.org): The central metadata registry for Investigations, Studies, Assays, and DataFiles.
2. WorkflowHub (workflows.cytognosis.org): RO-Crate registry for computational workflows.
3. LaminDB (Planned): Internal data artifact registry.
4. Zoekt (code.cytognosis.org): Internal fast code search index.
Data is stored in Google Cloud Storage (GCS) utilizing uniform bucket-level access.
gs://cytognosis-audit-7yr): 7-year immutable retention policy enforced via Cloud Storage Bucket Lock.us-central1.All data artifacts are tracked using a 4-layer provenance stack:
1. DVC: Content-addressed hashing for data files.
2. redun: Execution provenance for pipelines.
3. MLflow: Model and experiment tracking.
4. RO-Crate: Standardized packaging for publication.
Every asset registered in the VFS receives a Software Heritage Identifier (SWHID) and a W3C PROV-J sidecar file.