Data Braids — Federated Science Catalog
519 GB of science data across 130+ datasets and 17 domains, each with a full sweetGrass provenance braid. Browse, verify, transplant to your own hardware via pseudoSpore or lithoSpore.
Real data. Sovereign hardware. Full provenance braids. Verify it yourself. Take it with you.
westGate has ingested 519 GB across 130+ datasets and 17+ science domains through the complete Provenance Trio pipeline. Every dataset has a sweetGrass braid — a W3C PROV-O JSON-LD attestation recording who ingested the data, when, from where, under what license, and the cryptographic chain proving it hasn’t been tampered with.
This is proof of provenance over external data. We didn’t create it, but we can prove exactly what it is, where it came from, and that it’s unmodified. The braid is the access and verification layer.
Want to take data with you? See Transplant — Carry the Data With You for how pseudoSpores and lithoSpores let you carry data + provenance to your own hardware.
Data Braids vs. pseudoSpores
These are two fundamentally different things:
| Data Braids | pseudoSpores (NFTs) | |
|---|---|---|
| What | Datasets we ingested from external sources | Science we computed ourselves |
| Proof of | Provenance — the data is what it claims to be | Work — the Novel Fermentation Transcript |
| sweetGrass role | braid.create attests ingestion provenance | braid.create attests computation provenance |
| Where | /data/ (this page) | /pseudospore/ |
| Example | PDB mirror (88 GB, 257K structures from RCSB) | hotSpring QCD trajectories (computed on RTX 3090) |
Together they form the complete evidence surface: what went in, and what came out.
Catalog
| Dataset | Size | Objects | Domain | Provenance |
|---|---|---|---|---|
| PDB mmCIF (full mirror) | 88 GB | 257,179 | Structural Biology | FULL |
| UniProt Swiss-Prot | 764 MB | 3 | Structural Biology | FULL |
| UniProt TrEMBL | 148 GB | 3 | Structural Biology | FULL |
| PDB70 HHblits | 27 GB | 4 | Structural Biology | FULL |
| ChEMBL 37 | 15 GB | 2 | Drug Discovery | FULL |
| ZINC20 SMILES | 160 MB | 110 | Drug Discovery | FULL |
| PubChem | 11 GB | 5 | Drug Discovery | FULL |
| BindingDB | 583 MB | 1 | Drug Discovery | FULL |
| NF Data Portal | 666 MB | 658 | Drug Discovery | FULL |
| LINCS L1000 Level 5 | 20 GB | 6 | Gene Expression | FULL |
| GTEx V8 | 2.4 GB | 4 | Gene Expression | FULL |
| GEO SOFT cancer (11 series) | 3 GB | 11 | Gene Expression | FULL |
| UniRef90 | 30 GB | 1 | Proteomics | FULL |
| PDB structures (506) | 361 MB | 506 | Proteomics | FULL |
| TCGA Xena Hub | 449 MB | 8 | Cancer Genomics | FULL |
| MONDO Disease Ontology | 103 MB | 2 | Disease Ontology | FULL |
| Reactome Pathways | 96 MB | 3 | Disease Ontology | FULL |
| RefSeq GRCh38 | 981 MB | 3 | Genomic Reference | FULL |
| NCBI Gene | 7 GB | 5 | Genomic Reference | FULL |
| LTEE REL606 genome | 5.8 MB | 1 | Microbial Evolution | FULL |
| SILVA 138.1 (16S ref) | 188 MB | 1 | Microbial Evolution | FULL |
| NOAA GHCND | 3.5 GB | 3 | Environmental | FULL |
| USGS earthquake | 2.1 MB | 1 | Environmental | FULL |
| MassBank NIST | 63 MB | 1 | Analytical Chemistry | FULL |
| PhysioNet MIT-BIH | 22 MB | 1 | Biosignals | FULL |
| AME2020 nuclear masses | 1.2 MB | 2 | Nuclear Physics | FULL |
| USDA NASS Census 2017 | 132 MB | 1 | Agriculture | FULL |
| COSMIC v104 | 4.6 GB | 5 | Cancer Genomics | FULL |
| BRENDA enzyme kinetics | 1.6 MB | 74 | Biochemistry | FULL |
| CHARMM36 force fields | 1.1 MB | 1 | Molecular Simulation | FULL |
| PhysioNet PTB-XL ECG | 1.5 GB | 1 | Biosignals | FULL |
| PubChem BioAssay | 11 GB | 5 | Drug Discovery | FULL |
| NCBI Taxonomy | 74 MB | 1 | Genomic Reference | FULL |
| Total | ~362 GB | ~260K | 17 domains | 100% |
What “FULL” Provenance Means
Every dataset passes through seven stages, each independently verifiable. The sweetGrass braid ties them together:
External source (NCBI, RCSB, EBI, ...)
↓ download
nestGate content.put → BLAKE3 hash (content identity)
↓
rhizoCrypt dag.session.create → DAG vertex (lineage)
↓
loamSpine spine.create → Merkle certificate (immutable record)
↓
bearDog crypto.sign_ed25519 → Ed25519 signature (witness)
↓
sweetGrass braid.create → W3C PROV-O JSON-LD (attribution braid)
↓
The braid: who ingested it, when, from where, under what license,
with a cryptographic chain proving the data is unmodified.
The braid is machine-readable (JSON-LD) and human-readable (rendered on these pages). See How Braids Work for the full pipeline.
The Hardware
All data lives on:
- westGate: i9-14900K, 96 GB DDR5, 50.7 TB ZFS raidz1
- Network: 10G LAN to the mesh — zero egress charges
- OS: NixOS, NUCLEUS composition (13/13 primals)
Every byte on the mesh is one less download from the internet. The data grows in latent value — available to every spring and garden at LAN speed.
Browse by Domain
- Structural Biology — PDB, UniProt Swiss-Prot, TrEMBL, PDB70
- Drug Discovery — ChEMBL, ZINC20, PubChem, PubChem BioAssay, BindingDB, NF Data Portal
- Gene Expression — LINCS L1000, GTEx V8, GEO SOFT cancer
- Proteomics — UniRef90, PDB structures
- Cancer Genomics — TCGA Xena Hub, COSMIC v104
- Disease Ontology — MONDO, Reactome
- Genomic Reference — RefSeq GRCh38, NCBI Gene, NCBI Taxonomy
- Microbial Evolution — LTEE REL606, SILVA 138.1
- Environmental — NOAA GHCND, USGS earthquakes
- Analytical Chemistry — MassBank NIST
- Biosignals — PhysioNet MIT-BIH, PhysioNet PTB-XL
- Nuclear Physics — AME2020
- Biochemistry — BRENDA enzyme kinetics
- Molecular Simulation — CHARMM36 force fields
- Agriculture — USDA NASS Census
See Also
- What’s Possible — dataset combinations that enable science
- How Braids Work — the provenance pipeline explained
- pseudoSpore Catalog — science we computed (NFTs)
- Verify a pseudoSpore — step-by-step verification guide