BioSample

BioSample is an NCBI database that stores descriptions of the physical biological source material behind an experimental dataset — a cell line, a tissue biopsy, an environmental water sample, a bacterial isolate — kept as a distinct record from the sequence or experimental data itself. Where a sequencing read or a GenBank record answers “what was measured,” a BioSample record answers “what organism or material was it measured from, and under what conditions was it collected”: organism, strain, tissue type, developmental stage, geographic location, collection date, host, and dozens of other attributes depending on sample type. This separation matters in practice because the same physical sample commonly feeds multiple downstream datasets (a genome assembly, a set of RNA-seq runs, a variant call set), and BioSample is what lets all of them point back to one authoritative description of where the material actually came from.

Accessions and the package/attribute model

Every registered sample receives a SAMN-prefixed accession. Rather than a single free-text description field, BioSample enforces structured, consistent attribute names and values through a defined set of packages — a package is a template of the specific fields expected for a given sample type, so a human clinical isolate, an authenticated cancer cell line, and a soil metagenome sample are each described with the fields actually relevant to that kind of material, not one generic form. Environmental and genomic samples specifically use the MIGS/MIMS/MIMARKS checklist standards (Minimum Information about a Genome/Metagenome Sequence, and about a Marker Gene Sequence), a community standard that predates and feeds into BioSample’s own package system, which is why a metagenomics submission is expected to report host, isolation source, collection date and geographic coordinates as a matter of course rather than as optional extras.

How BioSample connects to the rest of NCBI’s archives

BioSample does not stand alone. A BioProject (accession prefix PRJNA) is the umbrella record for a research initiative or study, and typically links to many individual BioSample records underneath it — one BioProject, many samples. Each BioSample record in turn links out to whatever data was actually generated from that material: raw reads deposited in the Sequence Read Archive (SRX/SRR-prefixed experiment and run accessions), assembled or annotated sequence in GenBank, or, for human genotype-phenotype data requiring controlled access, a linked dbGaP record instead of a public one. All of these submission types run through NCBI’s shared Submission Portal, which performs metadata validation up front and assigns a temporary SUB-prefixed working ID before the permanent SAMN/PRJNA/SRX accessions are issued — meaning a BioSample submission is rarely a standalone action, but one step in registering a coordinated set of linked records for the same underlying study.

Why the sample/data separation is the actual design decision

The reason BioSample exists as a separate database, rather than folding sample description into each sequence record, is reuse: a single tissue sample or cell line often generates multiple, unrelated experiments over time (an initial genome assembly, then later RNA-seq, then later still a variant-calling reanalysis), and repeating the same provenance metadata in every downstream record would both bloat submissions and risk the copies drifting out of sync. Centralizing it in one BioSample record, referenced by accession from every downstream dataset, keeps that provenance authoritative and lets a researcher discover every dataset derived from a given sample by querying the BioSample accession itself, rather than reconstructing the relationship after the fact.

Where BioSample fits among named repositories

BioSample is infrastructure, not a destination dataset in its own right — researchers rarely cite “BioSample” alone the way they’d cite the GDC or gnomAD; instead they cite the specific SAMN accession(s) their downstream data depends on. A data management plan describing a wet-lab study that will deposit sequence data to a public archive should name the intended BioSample package and attribute set up front (organism, MIGS/MIMS/MIMARKS fields if applicable) rather than deferring sample metadata to submission time, the same accession-level specificity CASRAI’s own Data Management Plan guidance recommends for any named repository dependency.