Visualising specimens as event chains on the Australian Reference Genome Atlas

What this was about

Kathryn Hall described how the Australian Reference Genome Atlas (ARGA), an index rather than repository, modelled provenance chains from a wild organism through collection, museum registration, tissue subsampling, DNA extraction and lab runs to deposited genomic data products. Because Darwin Core is flat and genomic data derive from many nested material derivatives, ARGA built linked back-end tables keyed on organism IDs, an RDF-based ETL layer, a relational core with GraphQL access, a taxonomic DAG aligning ALA, WoRMS and NCBI taxonomies, and operation-based CRDTs recording every database action. The result lets users see, for example, 31 derived data products from one specimen with tissue, molecule, file type and custodian preserved.

Why it matters. Provenance is the 'data seatbelt' that matters when a sequence or phylogeny looks wrong, and ARGA shows a working model for linking specimens to genomic data at fine grain.

In the room

• ARGA was a partnership of the Atlas of Living Australia, Australian BioCommons and Bioplatforms Australia, with ARDC; the project concluded earlier this year.
• Genomic data come from amplicons from extracts from subsamples; one organism can yield hundreds of amplicons and thousands of fragments.
• Back-end tables (organism, live state, collections, registrations, tissues/lab samples, data products) are linked by organism ID.
• Organism ID tracks provenance even when tissues were never formally registered in collections.
• A taxonomic directed acyclic graph aligns ALA, WoRMS and NCBI taxonomies without flattening rank shifts.
• Operation-based conflict-free replicated data types atomise records and preserve a full history of actions.
• Aim was a sufficiency of metadata to reconstruct the chain, not every possible detail.

Notable moments