Datasets

Sources

dbSNP

eQTLGen

Open Targets

Pan-UK Biobank

UK Biobank Pharma Proteomics Project

Organization

Every dataset is identified by two levels — namespace, table — plus optional Hive partitioning at the physical layer:

ukb_ppp / pqtls / ancestry=EUR / protein_id=P43220 / data.parquet
└ namespace ┴ table └─────── partitions ────────┘

Level	Meaning	Example
Namespace	collection of related tables	ukb_ppp
Table	dataset	pqtls
Partition	physical shard (large tables only)	ancestry, protein_id

The same identifier shape maps to every catalog the data is exposed through:

System	Reference for ukb_ppp.pqtls
R2 Object Storage	s3://bedrock-bio-data/ukb_ppp/pqtls/
R2 Data Catalog (Iceberg)	bedrock-bio-data.ukb_ppp.pqtls

Parquet files in Cloudflare R2 are the single source of truth. R2 Data Catalog wraps them as Iceberg tables for the open-source clients.

Access

RThin DuckDB-backed client over the Iceberg tables.

PythonSame client pattern, Python edition.

Agent skillsEquip Claude Code, Codex, Antigravity, or OpenCode to check Bedrock Bio first for relevant public data.