Biology · Genomics, Omics & Bioinformatics
Biological databases, stable identifiers and data provenance
A fictional sequence changes from ACGT to AACGT after a leading position is added. The letters CGT now occupy different coordinates. A second update changes only the annotated interval on the original sequence. These are different revisions, even if a convenient search name looks unchanged.
Gene Expression: RNA & Proteins explains how sequence, RNA and protein are related. Phylogeny, Classification & the Tree of Life introduces ancestry and the interpretation of trees. Research Design, Data & Scientific Practice explains sampling, independent units and the limits of comparisons. Their English chapters are available in this working manuscript; publication remains pending. The examples here use letter strings, inclusive coordinates, averages, fractions and ordinary graph reading. Each convention is introduced before it is used.
Back to the Biology mapBy the end, you can
Identify the exact object, sequence version, annotation release and derivation history needed to interpret or recompute a biological data claim.
A name and a versioned object have different jobs
A familiar gene name is useful for communication but may refer ambiguously to an entity across organisms, databases or contexts. A database identifier names a defined record or feature. A version identifies a particular state under that database's rules. The object might be a sequence, gene, transcript, protein, assembly or sample; those identities should not be collapsed into one interchangeable label.
Stable identifiers aim to preserve continuity, but their version rules depend on what they represent. A sequence update and an annotation update need not trigger the same revision. Some systems also change assembly versions for changes in represented sequence names. The appropriate rule comes from the relevant resource, rather than from the assumption that every dotted number means the same kind of change.
The original history uses explicitly fictional Ark identifiers. Its local rule changes the sequence version when the sequence string changes, while annotation releases are recorded separately. It illustrates the distinction without imitating an actual accession or claiming that every real database follows this simplified rule.
Sequence and annotation revisions can differ
Read the diagram in words
ACGT positions 2–3 give CG; positions 2–4 give CGT under another annotation release. AACGT positions 3–5 retain CGT after a leading A is added. Old positions 2–4 would instead give ACG.
| Fictional record state | Exact sequence string and version | Separately recorded annotation release | Inclusive start | Inclusive end | Selected feature letters | Selected interval length |
|---|---|---|---|---|---|---|
| First represented feature | ACGT · ARK_TOY_SEQ.1 | toy-annotation-a | 2 | 3 | CG | 2 |
| Revised annotation on the same sequence | ACGT · ARK_TOY_SEQ.1 | toy-annotation-b | 2 | 4 | CGT | 3 |
| Revised sequence and shifted feature address | AACGT · ARK_TOY_SEQ.2 | toy-annotation-c | 3 | 5 | CGT | 3 |
Same sequence, changed interpretation
New sequence, changed address
Actual resources have documented entity-specific version rules
Fictional Ark identifiers and a deliberately specified local version rule: change sequence version when the string changes; track annotation releases separately. All intervals are one-based inclusive. No real accession, database retrieval, gene sequence or universal database rule.
Coordinates belong to a particular representation
A coordinate is an address within a named representation and convention. Positions two through four in ACGT are CGT, but the same letters occupy three through five in AACGT. Reusing the old coordinates against the new string would select a different piece. A bare interval therefore cannot identify a biological feature on its own.
The history also changes an interval from two/three to two/four while retaining the original sequence version. That is an annotation revision: the represented feature expands from CG to CGT. A later sequence revision shifts the CGT interval to three/five. Keeping both revision histories explains which change affected the letters and which affected the interpretation.
A citation or saved analysis can preserve the exact sequence version, annotation release, coordinate convention and relevant sample metadata. Retrieval timing is useful when a resource changes, but a date alone may not identify its exact contents. A persistent identifier also needs the appropriate object and version to support an exact comparison.
A derivation record explains how an output arose
Provenance describes the inputs, definitions and transformations behind a result. In the original graph, read records and metadata feed a stated assembly representation; an annotation release adds feature interpretations; an analysis produces a named summary. The arrows record derivation and interpretation, not a biological pathway.
A recomputation needs more than an output file. Changed reference versions, filtering choices or software settings can change results even with the same raw observations. Retaining those identities helps explain a discrepancy instead of silently treating every difference as a new biological discovery.
Access and scientific reliability are separate matters. Public availability can make a record easier to inspect without ensuring that its annotations are correct or that every intended reuse is appropriate. Sensitive sample metadata can require access conditions. Appropriate stewardship preserves useful provenance while respecting the conditions under which data may be shared.
A result retains its derivation account
Read the diagram in words
Read records and metadata inform the assembly representation. That representation supports an annotation release. The assembly, annotation and documented analysis choice together inform the summary.
| Represented record or transformation |
|---|
| Original read records |
| Sample identities and definitions |
| Exact assembly representation |
| Exact annotation release |
| Documented analysis choice |
| Named derived output |
| Represented record or transformation | Documented provenance relation |
|---|---|
| 1. Original read records → Exact assembly representation | Source data retained |
| 2. Sample identities and definitions → Exact assembly representation | Definitions needed to read the input |
| 3. Exact assembly representation → Exact annotation release | Exact representation and release |
| 4. Exact assembly representation → Named derived output | Result derived under those inputs and choices |
| 5. Exact annotation release → Named derived output | Exact representation and release |
| 6. Documented analysis choice → Named derived output | Choice controlling the computation |
Traceability does not prove correct biological inference
Original conceptual record lineage. Relations represent input, interpretation and derivation, not biological causation or an actual executed software pipeline. No external data retrieval, algorithm, operational genomic procedure or correctness guarantee.
Worked example
Keep the two revision histories distinct
In the first fictional record, sequence version one is ACGT and annotation release a identifies positions two through three, giving CG. Release b retains that same sequence but extends the interval through four, giving CGT. The sequence version does not change under the stipulated local rule.
The second sequence version is AACGT. To retain the CGT feature, annotation release c identifies positions three through five. Looking at positions two through four of the new string instead would give ACG.
An exact comparison therefore needs both the sequence identity/version and the annotation release with its coordinate convention. The provenance graph adds how those records were used; neither a common display name nor a final number alone reconstructs that account.
Check your understanding
Why can an annotation change occur without a sequence-version change in this model?
Show reasoning
The local rule tracks sequence strings separately from feature interpretations. Release b changes the interval on the same ACGT string. Actual databases have their own documented entity-specific rules.
What happens if positions two through four are reused after the leading A is added?
Show reasoning
They select ACG rather than CGT. The intended CGT occupies positions three through five in AACGT, so coordinates must stay tied to the exact representation and convention.
Does public access to the output establish a reproducible and scientifically reliable result?
Show reasoning
No. Exact inputs, versions and derivation choices are needed to recompute it. Even an exact recomputation does not establish valid sampling, correct annotation or appropriate reuse of sensitive information.
What to carry forward
Preserve the represented object, exact versions, coordinates and derivation; a searchable name is only the beginning of the evidence account.
References
Resources supporting this lesson’s explanations and examples.
- Sequence Identifiers (Historical page updated 9 November 2017)NCBI/NLM and INSDCOfficial data documentation
- Stable IDs (Ensembl release 116, June 2026 archive)Ensembl/EMBL-EBIOfficial archived data documentation
- Genome assembly versioning and status (Page generated 28 September 2026; inspected 13 October 2026)NCBI/NLMOfficial data documentation