Why Bioinformatics Runs on Databases

By the end of this lesson you will know why almost every bioinformatics task starts by asking a database a question, and you will have a concrete goal to work toward: pulling everything known about a single gene, BRCA1, out of a dozen separate repositories and stitching it into one clean record. This is the deep dive — the intro course lesson on biological databases showed you the front doors of NCBI, UniProt, and Ensembl; here we go behind them, and we do not stop until you can retrieve, cross-reference, and script against them by hand.

The data has outgrown every lab

A single modern sequencing run produces more raw text than a human could read in a lifetime. No individual lab can store, curate, or even see all of it. So the field made a trade: instead of each group hoarding its own files, everyone deposits into shared public archives and everyone draws back out.

That trade is why a database sits at the start of nearly every analysis. You almost never begin from a blank sequence. You begin from a record someone already deposited, checked, and gave a permanent name. An accession number is that permanent name: a stable, permanent ID for one database record, so you and a reader on the other side of the planet can fetch the exact same entry years apart. BRCA1's curated messenger-RNA record, for instance, is NM_007294 at NCBI, and it will always be that.

INSDC: three archives, one shared record

The three big nucleotide archives are not rivals. They are one system.

The INSDC (International Nucleotide Sequence Database Collaboration) binds together GenBank at NCBI in the United States, the ENA (European Nucleotide Archive) at EMBL-EBI, and the DDBJ (DNA Data Bank of Japan). Under this agreement they synchronise their holdings daily: submit a sequence to any one partner and, within about a day, the other two carry an identical copy under the same accession.

The practical payoff: an accession like NM_007294 is not "an NCBI thing" you have to reach through one specific website. It is a global address. This is also why later lessons can teach you to fetch the same record from whichever hub is fastest or least rate-limited for you.

FAIR: what "usable" actually means

Storing data is easy. Making it usable is the hard part, and the field has a name for the target: the FAIR principles — data should be Findable, Accessible, Interoperable, and Reusable.

Read those as four concrete demands you will lean on all course long:

• Findable — every record carries a unique, persistent ID (that accession again) so search can point straight at it.
• Accessible — you can retrieve it over a plain, documented protocol, by machine, not just by clicking.
• Interoperable — records speak in shared vocabularies and link out to each other, so BRCA1's gene record can name its own protein, structures, and papers.
• Reusable — enough provenance and licensing travels with the data that someone else can trust and reuse it.

Nearly everything you will do here is FAIR in action: you find BRCA1 by ID, access it by API, follow interoperable cross-links between databases, and produce a reusable record at the end.

A named skill, not a nice-to-have

This is not a side quest. The ISCB (International Society for Computational Biology) publishes the field's core competencies, and it lists the ability to retrieve, manipulate, and manage data from public repositories as a foundational one — right alongside programming and statistics. Employers assume it. This course is where you actually build it.

The promise: one gene, a dozen databases

Here is where we are headed. By the final lesson you will assemble a complete BRCA1 dossier from records you fetch yourself:

• NCBI Gene ID 672 — the gene-centric hub that names everything else
• RefSeq mRNA NM_007294 and protein NP_009225
• Ensembl gene ENSG00000012048
• UniProt P38398 (the reviewed BRCA1_HUMAN entry)
• KEGG hsa:672 for pathways
• PDB structures such as 1JNX (the BRCT domains) and 1JM7 (the BRCA1–BARD1 complex)

Twelve-plus IDs, twelve-plus databases, one gene. They connect because the records point at each other, and because tools like NCBI's ESearch let you query them programmatically — up to 3 req/s without an API key, or 10 req/s with one. Assembling that dossier by hand, cross-link by cross-link, is the project this whole course builds toward. Before we can fetch anything, though, we need to know what kind of record each ID names — so the next lesson, Primary vs Derived Databases, draws the first and most important line on the map.