GenBank and the Nucleotide Archive

By the end of this lesson you will be able to open any GenBank record, read it top to bottom — the LOCUS line, the FEATURES table, the ORIGIN sequence — and explain why the same gene shows up in the archive dozens of times. We anchor it all in the record where BRCA1 first entered public science.

What GenBank is — and what it deliberately is not

GenBank is the open, submitter-populated nucleotide archive. "Submitter-populated" is the whole personality of it: anyone who sequences DNA or RNA — a graduate student, a genome consortium, a diagnostics lab — deposits their sequence, and it becomes a permanent public record under their name. NCBI stores and serves it, but does not rewrite it. The record belongs to whoever submitted it.

GenBank is one third of the INSDC — the International Nucleotide Sequence Database Collaboration. The other two members are ENA (the European Nucleotide Archive at EMBL-EBI) and DDBJ (the DNA Data Bank of Japan). The three exchange new records every day, so a sequence deposited in any one appears in all three within about a day. When the two great hubs lesson said NCBI and EMBL-EBI mirror each other for sequence data, INSDC is the machinery that makes it true. Deposit once, and the world's three archives hold identical copies.

This makes GenBank a primary, archival database in exactly the sense of the primary-vs-derived lesson earlier in this course: it records what was observed and submitted, warts and all, and it never edits that observation out from under you.

Anatomy of a GenBank flat file

BRCA1 entered public databases in 1994, when Miki and colleagues cloned it. Their submission is U14680 — a real, frozen GenBank record you can pull today. Its flat file (the plain-text .gb format) is divided into labelled sections that always appear in the same order:

LOCUS       HSU14680        5711 bp    mRNA    linear   PRI 10-JUN-2002
DEFINITION  Homo sapiens breast and ovarian cancer susceptibility
            (BRCA1) mRNA, complete cds.
ACCESSION   U14680
VERSION     U14680.1
SOURCE      Homo sapiens (human)
REFERENCE   1  (bases 1 to 5711)
  AUTHORS   Miki,Y., Swensen,J., ... Skolnick,M.H.
  TITLE     A strong candidate for the breast and ovarian cancer
            susceptibility gene BRCA1
FEATURES             Location/Qualifiers
     ...
ORIGIN
        1 gctgagactt cctggacggg ggacaggctg tggggtttct cagataactg ...

Reading top to bottom: LOCUS is the one-line summary — an internal name, the length (5711 bp), the molecule type (mRNA), topology (linear), a three-letter division code (PRI), and the last-modified date. DEFINITION is the human-readable description. ACCESSION is the stable ID; VERSION adds the .version suffix you met in the versioning lesson — here U14680.1, edited zero times since deposit. REFERENCE links the sequence to its publication. ORIGIN is the sequence itself, in blocks of ten bases with position numbers down the left.

The division code sorts records by broad category: PRI for primate, ROD for rodent, BCT for bacterial, VRL for viral, EST for expressed-sequence tags, PAT for patented sequences, and more. It is a legacy filing system, not deep biology, but it tells you at a glance what kind of thing you are looking at.

The FEATURES table: where the biology lives

The richest section is FEATURES — a structured annotation of what the sequence means, laid out as feature keys (like source, gene, CDS) each with a location and a set of /qualifier="value" lines:

     source          1..5711
                     /organism="Homo sapiens"
                     /mol_type="mRNA"
                     /db_xref="taxon:9606"
                     /chromosome="17"
     gene            1..5711
                     /gene="BRCA1"
     CDS             120..5711
                     /gene="BRCA1"
                     /codon_start=1
                     /product="breast and ovarian cancer susceptibility"
                     /protein_id="AAA73985.1"
                     /translation="MDLSALRVEEVQNVINAMQKILECPICLELIKEPVSTKCDHIFC..."

Two things to notice. The CDS (coding sequence) feature carries a /translation — the protein GenBank derives by reading the RNA from position 120 to 5711 — and a /protein_id, AAA73985.1, its own accession in the protein archive. And that /db_xref="taxon:9606" is a cross-reference: a pointer from this record into another database (here NCBI Taxonomy, where 9606 is Homo sapiens). Records point at each other constantly through db_xref, a mechanism a later lesson is devoted to.

You rarely read a flat file by eye. Biopython parses every section into attributes:

from Bio import Entrez, SeqIO

Entrez.email = "[email protected]"          # NCBI asks who you are
with Entrez.efetch(db="nuccore", id="U14680.1",
                   rettype="gb", retmode="text") as handle:
    rec = SeqIO.read(handle, "genbank")

print(rec.id, len(rec.seq), "bp")         # U14680.1 5711 bp
for feat in rec.features:
    if feat.type == "CDS":
        print(feat.qualifiers["protein_id"])   # ['AAA73985.1']

Keep your request rate polite — E-utilities calls are capped at 3 req/s without an API key, or 10 req/s with one registered to your NCBI account.

Redundancy: why one gene fills many records

Because GenBank accepts every submission and never merges them, BRCA1 is not one record — it is hundreds. U14680 is the 1994 mRNA. Other labs deposited their own BRCA1 clones, partial exons, and variant sequences; U14680 itself carries a /note pointing at U15595 for an alternatively spliced exon. Each is a legitimate, independent observation. None is "the" BRCA1 sequence.

That is the natural next step: the same BRCA1, seen not through the raw archive but through a curated lens. The RefSeq lesson takes U14680's messy, many-record world and shows you the single reviewed NM_007294 that stands in for it.