Protean just wired a real dataset lake into the runtime.
APD/APD6, DRAMP, PeptideAtlas, UniProt Swiss-Prot, ChEMBL targets, and MedQA are now visible through a governed readiness layer.
Not “throw data at a model.”
The system now checks what exists, what has provenance, what has checksums, what is blocked by licensing, and what is allowed only as evidence/context.
That matters for bringing DeSci to
@base.
If the goal is token → experiments → equity pathways, the rails need to know the difference between raw data, reviewed labels, and proof.
More data is easy.
Accountable data is the hard part.