Why Science? | European Nucleotide Archive Study, Sample and Run Evidence

Sequence data can move through several named objects before becoming a FASTQ file on a computer. The European Nucleotide Archive, or ENA, preserves studies, biological samples, sequencing experiments, runs and downloadable files within the international nucleotide-data system.

That hierarchy is the key to responsible reuse. A run identifies a sequencing execution; it is not automatically a separate organism or biological replicate. A sample records biological material; it does not by itself describe library construction. This guide keeps those layers connected.

Build an accession map, sample table and file manifest before alignment or assembly. The fastest pipeline cannot repair a file assigned to the wrong organism, condition or library.

Define the intended analysis before choosing a file type. A transcript-expression question, microbial amplicon survey, genome assembly and variant analysis require different library designs and quality checks. The archive supplies sequence objects; suitability comes from matching those objects to the question.

Start with a small study and preserve the original metadata export. Create a read-only source directory, a manifest with accessions and checksums, and a derived directory for cleaned metadata and analysis output. This layout keeps every transformation reversible.

Read the hierarchy as an experimental design, not a download catalogue. The study explains scope, the sample describes material, the experiment records library preparation and the run points to sequencing output. Skipping a level removes information that downstream software cannot reconstruct reliably.

Sampling context matters for environmental and clinical sequences. Collection site, host, tissue, date, treatment and extraction details can shape composition. Record the fields needed for the comparison and flag inconsistent units or naming before grouping samples.

Contamination and mixed organisms require cautious identity language. A taxon field may describe intended source, while reads can contain adapters, host sequence or environmental mixtures. Use read-level quality and classification methods for sequence-content claims rather than treating metadata as a measurement.

Reference genomes and annotations add another versioned layer. Record assembly accession, release and indexing parameters when aligning reads. Results from the same FASTQ files can differ when reference content or transcript definitions change.

Quality control should be planned before filtering. Preserve original read counts, adapter decisions, trimming thresholds and post-filter counts per run. A high percentage of retained reads is not automatically good if the starting files belong to the wrong library or sample.

For paired runs, maintain mate relationships through decompression, renaming and filtering. Use stable run accessions in filenames and tables. Sorting files alphabetically without explicit pairing rules can join mates from different samples.

When combining runs, justify the merge using shared sample and compatible experiment metadata. Record the run list and keep per-run quality reports. A merged file may simplify software input while hiding a lane-specific or batch-specific problem.

The final report should include an input manifest, software versions, reference identifiers, sample exclusions and a count audit. These details let another reader distinguish biological findings from choices made during retrieval and processing.

Before download, compare expected sample, experiment and run counts with the archive table. A difference may reflect multiple lanes, repeated sequencing or an incomplete filter. Resolve it while the relationships are visible instead of after files have been renamed.

Preserve archive filenames or add accessions without deleting the originals. Human-friendly labels can live in a mapping table. Renaming alone is fragile because two projects may contain identical generic filenames such as reads_1.fastq.gz.

Compression is a storage layer, not a biological transformation. Record whether software reads compressed files directly or whether decompression creates temporary copies. Verify checksums at the stage for which the archive supplies them.

Read counts and base counts provide useful inventory checks, while quality profiles require inspection of the sequence files. Neither metric proves correct sample identity. Use them together with the accession map rather than as substitutes for it.

Adapters and low-quality tails may need trimming. Save the adapter source, software version, thresholds and counts before and after filtering. Aggressive trimming can remove real sequence or change comparability across runs.

For amplicon data, primer sequences and targeted region define what organisms or variants can be observed. Keep primer and protocol context beside the run; a FASTQ file alone cannot reveal the complete biological sampling frame.

For RNA sequencing, strandedness and library selection affect quantification. Confirm these properties from experiment metadata or methods instead of relying on an aligner's automatic guess. Incorrect orientation can produce plausible but misleading counts.

For whole-genome work, coverage varies across samples and genomic regions. Report depth and missingness rather than treating every unobserved variant as reference. Technical absence and biological absence must remain separate.

Assemblies and annotations are derived products with their own accessions and versions. If the project uses them, record how they connect to the raw runs. Do not cite an assembly accession as if it identified every input read file.

Environmental samples may have rich geographic or ecological fields. Preserve units, coordinate conventions and collection dates, and avoid publishing sensitive locations when the record or project requires responsible handling.

Human sequence data can involve controlled-access conditions in other systems even when metadata are discoverable. Follow archive access rules and study consent boundaries. Public metadata is not permission to make personal or clinical claims.

When API queries support a report, save the response schema and requested fields. A service may add columns or update records later. Keeping the raw response makes changes detectable and lets parsers be tested.

Joining several studies requires harmonisation across assemblies, library strategies and sample vocabularies. Keep each study's source table intact and add a cross-study mapping layer. Do not flatten incompatible designs into one comparison merely because all files are FASTQ.

A claim ledger can connect each statement to study, samples, runs, processing, reference and limitation. Review it before publication. If the statement has no exact run set, the evidence trail is too broad for reproducible reuse.

For peer checking, select one plotted point or assembled sequence and trace it back to run, experiment and sample. This reverse route often catches swapped labels that forward pipeline logs do not reveal.

Negative controls and blanks should remain visible when the study includes them. They can reveal reagent or environmental contamination, but they are not ordinary biological samples. Explain how their reads were used in filtering or interpretation.

Deduplicate manifests by accession and file identity, not by sample title. Two runs can share a descriptive name while representing different acquisitions, and mirrored download paths can refer to the same archive object.

If a run is withdrawn or unavailable, record that status and the retrieval date. Do not replace it silently with a neighbouring run. The missing object may change sample balance or prevent exact reproduction.

Use sufficient computational resources without allowing convenience to redefine the question. Subsampling can support teaching or pipeline tests, but save the random seed, method and selected reads and avoid presenting the subset as the full archive run.

Visual reports should label accessions, sample groups and processing stage. A quality plot without file identity cannot be tied back to the archive, while a phylogenetic tree without reference and filtering details cannot explain its evidence boundary.

These checks turn the accession hierarchy into a practical scientific method: identify, map, verify, process, count and only then interpret.

Store the completed accession map with the analysis code and final figures. Future updates can then change software or reference versions while preserving the exact biological inputs, making comparison deliberate instead of accidental.

That durable link from sample to sequence is the central evidence skill this guide develops.

Before analysis, capture the study or project accession, sample accessions, experiment accessions, run accessions, library strategy and file manifest. Verify that every downloaded file matches one intended run and that the run maps back to the correct biological sample.

Finish with counts and checksums: expected studies, samples, experiments, runs and files; downloaded objects; missing or duplicate links; and checksum results. Sequence pipelines can succeed on the wrong files, so identity checks belong before alignment or assembly.

Reading object	What it can support	Question to keep visible
Study or project	Overall research context and submitted collection	Which question and dataset scope does it define?
Sample	The biological material and its attributes	What organism, source and condition produced the sequence?
Experiment	Library construction, platform and sequencing design	How was this sample prepared and measured?
Run and files	A sequencing execution and downloadable data objects	Which files, formats and checksums belong to this run?

Sources and next reading

Substantive resource claims in this article were checked against the primary pages below. A database record is a structured route into evidence; it does not by itself prove a new biological conclusion. Recheck the live record, evidence labels and release context before using it in assessed work or research.

Primary sources

• ENA Browser and archive scope — accessed 24 October 2026.
• ENA submission guide — accessed 24 October 2026.
• ENA advanced search — accessed 24 October 2026.
• ENA example study record and accession hierarchy — accessed 24 October 2026.
• ENA Portal API documentation — accessed 24 October 2026.

Continue through eduKate

• GEO Series, Samples and functional genomics evidence
• BioStudies and study-level data packages
• Alliance of Genome Resources and cross-species gene evidence
• Science Learning Hub
• Education Hub
• Careers by Subject and Capability