The Pancreatic β-Cell Consortium (PBCC) uses DERIVA to archive, curate, and share heterogeneous data on β-cell biology and whole-cell modelling, ranging from soft-X-ray tomography and mass-spectrometry to computational models. DERIVA’s model-driven interface lets bench scientists, not DBAs, evolve the schema as new assays appear.
Why PBCC chose DERIVA
| Requirement | DERIVA capability |
|---|---|
| Rapid schema evolution as new imaging and omics assays are added | Scientists edit the ER model directly; Chaise UI and APIs update automatically. |
| Self-curation by labs across several institutions | Web forms enforce controlled vocabularies; role-based access keeps embargoed data private until release. |
| Link raw data → processed data → whole-cell models | Versioned object store (Hatrac) + provenance tables track every transformation and parameter set. |
| FAIR dissemination to the wider diabetes community | DOIs minted on release; landing pages expose rich JSON-LD for Google Dataset Search. |
Impact so far
- Six data types (SXT, fluorescence imaging, proteomics, transcriptomics, model meshes, simulation outputs) stored under one evolving schema.
- Schema refactor completed by a domain scientist—195 notebook cells, 65 DDL statements, zero DBA time—demonstrating that biologists can maintain complex relational models themselves.
- Public portal (data.betacell.isrd.isi.edu) serves curated datasets to modelling groups exploring β-cell electrophysiology and glucose-stimulated insulin secretion.
Typical workflow
- Lab uploads raw files plus minimal metadata via
deriva-cli; QC checks fire automatically. - Curator refines metadata in Chaise, linking specimens to protocols and controlled vocabularies (NCBI Taxon, MGI gene IDs).
- Data scientist executes a Jupyter notebook that generates processed volumes or whole-cell meshes and pushes results back as a new
ProcessedDataset. - Release toggles ACLs, assigns DOIs, and publishes the record to the public PBCC portal.
Lessons for other cell-modelling projects
- Schema-agnostic tools cut the lag between experimental innovation and repository support.
- Executable notebooks provide both the evolution script and provenance for review, making database change transparent to collaborators and funders.
- Continuous FAIRness—ingest-time identifiers and provenance—avoids costly post-hoc curation when datasets reach terabyte scale.
Reference: R.E. Schuler et al., “Database Evolution by Scientists, for Scientists: A Case Study,” Proc. IEEE e-Science 2023.