The Two Hubs: NCBI Entrez and EMBL-EBI

Almost every database you will ever touch lives inside one of two great institutions: the NCBI in the United States and EMBL-EBI in Europe. This lesson lets you place any resource — GenBank, Ensembl, UniProt, PDB — under the right roof, and understand the one organizing idea both hubs share even though they present it in opposite ways.

The Same Idea You Already Know

In Primary vs Derived Databases you learned the core split: primary (also called archival) databases store exactly what a submitter deposited and never silently change it, while derived (curated) databases add expert judgement on top of that raw material. Both hubs are built on this same distinction. What differs is how visibly each one draws the line — EMBL-EBI names the two halves out loud, and NCBI hides the seam behind a single search box.

EMBL-EBI: Deposition Databases vs Knowledgebases

EMBL-EBI (the European Bioinformatics Institute, an outstation of the European Molecular Biology Laboratory) splits its roughly forty resources into two named tiers.

A deposition database is a raw public archive — the place a scientist submits data straight from an instrument, with a permanent accession returned as a receipt. An accession number is a stable, permanent ID for one database record. Deposition databases include ENA (the European Nucleotide Archive, for sequencing reads and assembled sequence), PRIDE (mass-spectrometry proteomics), and MetaboLights (metabolomics). Nothing is second-guessed here; the archive's job is to preserve the record as submitted.

A knowledgebase is a curated layer built on top of those archives — a biologist reads the raw evidence, resolves conflicts, and writes an integrated record. Ensembl (genome annotation), UniProt (protein sequence and function), InterPro (protein families and domains), PDBe (3D structures), and ChEMBL (bioactive molecules) are knowledgebases. When you read BRCA1's UniProt entry P38398, you are reading a knowledgebase: a human curator merged the raw protein evidence into one authoritative summary.

NCBI: One Entrez System Over ~38 Databases

The NCBI (the US National Center for Biotechnology Information) makes the opposite design choice. It federates roughly 38 databases under a single integrated retrieval system called Entrez — one query language, one set of tools, one web search box that reaches across all of them at once.

The magic of Entrez is not that the databases sit side by side, but that their records are wired together by computed biological relationships. NCBI pre-calculates links so a record in one database points at related records in another: a nucleotide sequence links to the paper that described it, a protein links back to the coding sequence (CDS) that encodes it, and a protein links out to any solved 3D structure. You do not compute these joins — Entrez ships them.

So the primary/derived line still exists at NCBI; it just runs between member databases rather than being announced as a tier. GenBank is the archival deposition database (raw submitted sequence); RefSeq is the curated knowledgebase built from it. Both are reachable through the same Entrez search.

Hub to Flagship Resources

These are the same molecule seen from two buildings. BRCA1 is 672 and NM_007294 at NCBI, and ENSG00000012048 and P38398 at EMBL-EBI. Because both hubs mirror the international sequence archives (INSDC), the underlying data agrees — the wrappers differ.

Following BRCA1 Across Both Hubs

Entrez's computed links are directly scriptable through ELink, part of the E-utilities. This walks from BRCA1's Gene record to its protein records, then to any 3D structure, without you knowing a single structure ID in advance.

from Bio import Entrez
Entrez.email = "[email protected]"   # NCBI requires an identifying email

# Gene 672 (BRCA1) -> protein records, via a precomputed Entrez link
prot = Entrez.read(Entrez.elink(dbfrom="gene", db="protein", id="672"))
prot_ids = [l["Id"] for l in prot[0]["LinkSetDb"][0]["Link"]]

# Gene 672 -> solved 3D structures (NCBI Structure / MMDB)
struct = Entrez.read(Entrez.elink(dbfrom="gene", db="structure", id="672"))
print("protein records:", len(prot_ids))
print("structure links:", struct[0]["LinkSetDb"] != [])

At EMBL-EBI the same structure arrives by cross-reference instead: UniProt P38398 lists its PDB entries, such as the BRCT-domain crystal 1JNX and the BRCA1/BARD1 RING heterodimer 1JM7, which you can then open at PDBe. Same endpoint biology, two retrieval philosophies.

You now know which hub owns which resource. Next, A Field Map of the Major Databases lays every major resource out on one page so you can see both hubs side by side.