2 Biological Databases and Data Retrieval

What biological databases contain

Biological databases organize experimental observations, molecular sequences, structures, and curated knowledge so researchers can search, compare, and reuse them. A record typically connects biological data to an organism, sample, publication, annotation, and other database records, rather than consisting of a sequence alone.

Nucleotide sequence databases

The International Nucleotide Sequence Database Collaboration (INSDC) links GenBank at NCBI, the European Nucleotide Archive (ENA) at EMBL-EBI, and the DNA Data Bank of Japan (DDBJ). The partners exchange and synchronize submitted nucleotide records, so a sequence submitted to one is generally available through the others. These archives preserve submitted sequence data and associated information.

RefSeq differs from these archives: it is a non-redundant collection of reference sequences drawn from public sequence data, with varying levels of validation and curation. It includes genomes, transcripts, and proteins. RefSeq accessions commonly use prefixes such as NC_ for chromosomes or genomic molecules, NM_ for curated mRNAs, and NP_ for proteins; model-predicted transcripts and proteins often use XM_ and XP_. A prefix indicates record-type and status conventions, not a guarantee that every annotation is experimentally confirmed.

Raw sequencing reads are commonly archived in the Sequence Read Archive (SRA), while assembled or annotated nucleotide sequences are found in resources such as GenBank and ENA. These record types are not interchangeable: a read archive record is not the same as a reference genome or a gene transcript.

Protein sequence and annotation

UniProtKB provides protein sequences and functional annotation. Its two principal sections differ in review approach: Swiss-Prot is manually reviewed, while TrEMBL is computationally annotated and is not manually reviewed in the same way.

Use a UniProt accession as the stable identifier when citing an entry. The mnemonic entry name is useful, but is not the preferred stable identifier. Entries can include protein names, organism, sequence, function, domains and sites, subcellular location, evidence, references, and links to related records.

Macromolecular structures

The Protein Data Bank (PDB), maintained as a global archive by the Worldwide Protein Data Bank, stores three-dimensional structures of biological macromolecules. An entry can describe experimental methods, sample and polymer entities, chains, ligands, coordinates, and validation information. The entry identifier refers to the deposited structure record; chain identifiers distinguish molecular instances within that record. One protein may have several structures, and one structure may contain several chains or molecules.

PDB structures are distributed in PDBx/mmCIF and, for many entries, the older PDB format. For new computational work, check the entry’s current file-format options and choose a format supported by the analysis software.

Pathway knowledgebases

KEGG PATHWAY provides pathway maps for metabolism, signaling, and other biological processes. KEGG identifiers distinguish pathway maps, genes, ortholog groups, compounds, reactions, and other objects. For example, map00010 identifies a reference glycolysis/gluconeogenesis map, while hsa00010 is an organism-specific map for humans.

Reactome is another curated pathway knowledgebase, representing reactions, pathways, and the molecules involved. Its stable identifiers encode species and a database identifier; for example, an identifier beginning R-HSA- refers to a human event or entity.

Pathway records represent organized biological knowledge, not a direct measurement that every listed reaction is active in a particular sample. Experimental data can be mapped onto pathways, but interpretation requires attention to organism, evidence, and experimental context.

Identifiers, versions, and record fields

A database identifier is meaningful only together with its database and record type. A gene symbol such as TP53 is human-readable but may be ambiguous, species-dependent, or changed over time. A stable accession or numeric identifier is generally better for reproducible retrieval; record the database name and, when applicable, the identifier’s version.

For sequence records, an accession identifies the record. An accession with a version suffix, such as AB123456.2, identifies a particular sequence version. The accession can remain the same while annotation changes, whereas a sequence-version change means the sequence itself has changed. RefSeq accessions commonly include a version suffix as well. Use the exact accession and version required by the analysis or publication.

A typical annotated nucleotide record may include:

• Accession and version: record identity and sequence version.

• Definition and organism: a brief description and taxonomic source.

• References and submitter information: provenance and related publications.

• Features: annotated locations such as genes, coding sequences (CDS), exons, and regulatory regions.

• Sequence: the nucleotide string, often displayed in FASTA or a richer flat-file format.

• Cross-references: links to related gene, protein, assembly, or literature records.

FASTA is convenient when only the sequence and a short description are needed. A flat-file or XML record preserves more annotation and provenance. FASTA headers often include an accession, but the full header text is not a substitute for checking the original database record.

Finding and retrieving records

Use this workflow to make retrieval specific and reproducible:

1. Define the data needed. Decide whether the task requires a genomic region, transcript, protein, raw reads, structure, or pathway record.

2. Choose the appropriate database. For example, use RefSeq for a reference transcript, UniProtKB for protein-centered annotation, or PDB for a three-dimensional structure.

3. Search with specific terms. Combine a name or accession with organism, gene, molecule type, or other filters. A name alone may return records from multiple species or experiments.

4. Verify the record. Check the organism, sequence or entity, accession version, annotation status, evidence, and links to related records.

5. Select an output format. Use FASTA for sequence-only tasks; use a GenBank/EMBL flat file or XML when features and metadata matter; use PDBx/mmCIF or another suitable structure format for structural analysis.

6. Save provenance. Record the database, accession including version, retrieval date, query or filters, and file format. Database entries can be updated.

NCBI’s Entrez interface supports web searching across linked databases. For automated retrieval, NCBI E-utilities provide structured search and fetch operations: ESearch finds records matching a query, and EFetch retrieves records in a requested format. ENA’s Browser API supports direct retrieval by accession in formats such as FASTA, EMBL flat file, or XML. RCSB PDB supports searching and downloading structure files by entry identifier. These services can retrieve a known record or build a reproducible collection from defined search criteria.

For a nucleotide sequence, first identify the correct organism and record in NCBI Nucleotide or ENA, then retrieve it as FASTA if only sequence is needed. To obtain annotation as well, retrieve the richer flat-file or XML record. For a structure, search the PDB by accession, protein name, or sequence, then confirm the chain and experimental details before downloading coordinates.

Avoiding retrieval pitfalls

Several common errors can undermine reliable retrieval and interpretation:

• Confusing a gene symbol with a unique identifier: symbols can be shared or revised, so include the organism and use stable identifiers where possible.

• Mixing records for different objects: a gene, transcript, protein, assembly, and structure may each have distinct identifiers.

• Ignoring versions: an accession without its version may resolve to a later sequence than the one originally analyzed.

• Assuming equal evidence: archival submissions, computational predictions, and manually reviewed entries have different provenance and annotation depth.

• Treating formats as equivalent: FASTA omits most feature annotation, while flat-file, XML, and structure formats contain different information.

• Overlooking structure chains and biological context: a PDB entry may contain multiple chains, ligands, or copies of a molecule; confirm which entity the analysis concerns.

Choosing resources for the task

The resources serve different purposes: INSDC partners archive nucleotide data, RefSeq supplies reference sequences, UniProtKB organizes protein sequence and function, the PDB archives macromolecular structures, and KEGG and Reactome represent pathway knowledge. Reliable retrieval depends on choosing the right resource, identifying the exact record with a stable identifier and version, checking organism and evidence, selecting an appropriate format, and preserving retrieval details for reproducibility.