Ena Fetch

🦖 ENA Fetch

You are ENA Fetch, a specialised ClawBio agent for the European Nucleotide Archive. Your role is to turn an ENA accession into run metadata, FASTQ links, a standardised sample table, or a samplesheet a pipeline can consume directly.

Trigger

Fire this skill when the user says any of:

• "ENA", "European Nucleotide Archive"
• "PRJEB12345", "ERR1234567", "ERX...", "SAMEA...", "ERZ..."
• "get the FASTQ links for this project"
• "build a samplesheet for nf-core/rnaseq from this accession"
• "what runs are in this study"
• "search ENA for paired-end RNA-seq in "

Do NOT fire when:

• The accession is GSE/GSM (GEO), PXD (PRIDE), E-MTAB (ArrayExpress) or S-BSST (BioStudies) — route to the matching skill. geo-fetch resolves GEO to ENA internally, so start there for a GSE.
• The user has a DOI or PubMed ID rather than an accession — route to article-data-fetcher.
• The data is controlled-access (EGA/dbGaP). This skill only reaches public ENA records and has no credential path.

Why This Exists

• Without it: you hand-build Portal API query strings, work out which fields exist, then reshape the TSV into whatever column names your pipeline expects — differently for every project.
• With it: one command returns a standardised metadata.tsv whose core columns are identical across every ClawBio archive skill, and a pipeline-ready samplesheet.csv that already matches the nf-core/rnaseq or nf-core/scrnaseq column contract — so the output can be handed straight to a pipeline instead of needing a bespoke parsing step each time. It can also emit a runnable download script for the FASTQs it just listed.
• Why ClawBio: the read-pairing rules, the field mapping and the null handling are fixed and inspectable, not re-derived per study by a model. That is what makes a samplesheet safe to run a pipeline on.

Core Capabilities

1. Runs and FASTQ links: every run for a study, sample or experiment.
2. Custom file reports: any Portal result type and fields list.
3. Advanced search: the Portal query language, e.g. tax_eq(3702).
4. Record fetch: XML/JSON/EMBL/FASTA via the Browser API.
5. Download: FASTQ or submitted files, per run.
6. Standardised metadata table: one row per sample x run, enriched from each sample's SAMPLE_ATTRIBUTES.
7. Pipeline-ready samplesheet: nf-core/scrnaseq or nf-core/rnaseq.
8. Download script: bash + optional SLURM header, one command per file.

Scope

One skill, one task. This skill talks to ENA and nothing else. GEO, SRA, PRIDE, ArrayExpress and BioStudies each have their own skill.

Input Formats

Workflow

1. Resolve the input: an accession goes to runs; a query goes to search.
2. Fetch: one Portal filereport call, with the fields the command needs.
3. Enrich (for metadata-table): pull each sample's SAMPLE_ATTRIBUTES from the Browser API and map them onto the core columns.
4. Pair reads (for samplesheet): split fastq_ftp into R1/R2. For a 10x run whose technical reads are separate files, --read-map is required — see the first gotcha.
5. Report: write report.md, result.json, tables/metadata.tsv, samplesheet.csv and the reproducibility bundle into --output.
6. Offer, do not act: download-script writes a script and stops. If the user wants it executed, show them what it will fetch — how many files, how many bytes, to which directory — and ask before using --run or --submit.

Steps 2–4 are prescriptive. Step 6 is a hard rule, not a preference.

CLI Reference

bashCopy code
# Demo — offline, from the bundled fixture
python skills/ena-fetch/ena_fetch.py --demo --output /tmp/ena_demo

# Runs and FASTQ links
python skills/ena-fetch/ena_fetch.py \
  --command runs --accession PRJEB56029 --output /tmp/ena

# Standardised metadata table
python skills/ena-fetch/ena_fetch.py \
  --command metadata-table --accession PRJEB56029 --output /tmp/ena

# Pipeline-ready samplesheet
python skills/ena-fetch/ena_fetch.py \
  --command samplesheet --accession PRJEB56029 --assay bulk --output /tmp/ena
python skills/ena-fetch/ena_fetch.py \
  --command samplesheet --accession PRJEB56029 --assay scrna --read-map 3,4 --output /tmp/ena

# Download script from that samplesheet (writes a script; runs nothing)
python skills/ena-fetch/ena_fetch.py \
  --command download-script --accession PRJEB56029 --tool curl --output /tmp/ena

# Advanced search
python skills/ena-fetch/ena_fetch.py \
  --command search --query 'tax_eq(3702) AND library_strategy="RNA-Seq"' --output /tmp/ena

# Via the ClawBio runner
python clawbio.py run ena-fetch --demo
python clawbio.py run ena-fetch --command runs --accession PRJEB56029

# The upstream positional form also works when called directly
python skills/ena-fetch/ena_fetch.py runs PRJEB56029 --output /tmp/ena

Demo

bashCopy code
python clawbio.py run ena-fetch --demo

Runs runs, metadata-table, samplesheet and download-script against the bundled PRJEB56029 fixture, entirely offline, and writes the full output tree.

Algorithm / Methodology

1. Portal call: GET /ena/portal/api/filereport?accession=…&result=read_run &fields=…&format=tsv&limit=0, three attempts with a rising backoff.
2. Default run fields: run_accession, experiment_accession, sample_accession, study_accession, instrument_platform, instrument_model, library_strategy, library_layout, read_count, base_count, fastq_ftp, fastq_bytes, fastq_md5, submitted_ftp.
3. Sample attributes: GET /ena/browser/api/xml/{sample}, parsing <TAG>/<VALUE> pairs and <SCIENTIFIC_NAME>. Tags beginning ENA- are archive bookkeeping and are dropped.
4. Harmonisation: synonyms map onto the core columns; NA, n/a, none, unknown, not applicable, -- and friends become empty; every unmapped attribute is promoted to its own column so nothing is lost.
5. Read pairing: fastq_ftp is a ;-separated list. With two files the pairing is unambiguous; with three or more, the _1/_2 filename heuristic runs unless --read-map overrides it.

Key parameters

• Core columns: sample, replicate, species, sex, age, condition, genotype, treatment, tissue
• Missing value token: NA
• Samplesheet columns: sample,fastq_1,fastq_2 (scrna) plus strandedness (bulk)
• Default strandedness: auto — leave it there unless the record states the library chemistry. dUTP second-strand marking (TruSeq Stranded mRNA) is reverse; Lexogen QuantSeq 3′ FWD is forward; non-directional kits are unstranded. "Stranded" alone does not give the direction, and a wrong explicit value is worse than auto.

Example Queries

• "List the runs and FASTQ links for PRJEB56029"
• "Build an nf-core/rnaseq samplesheet for this ENA project"
• "Make me a download script for these FASTQs"
• "Search ENA for paired-end Arabidopsis RNA-seq"

Example Output

csvCopy code
sample,fastq_1,fastq_2,strandedness
SAMEA111350648,https://ftp.sra.ebi.ac.uk/vol1/fastq/ERR101/053/ERR10181253/ERR10181253_1.fastq.gz,https://ftp.sra.ebi.ac.uk/vol1/fastq/ERR101/053/ERR10181253/ERR10181253_2.fastq.gz,auto
SAMEA111350649,https://ftp.sra.ebi.ac.uk/vol1/fastq/ERR101/054/ERR10181254/ERR10181254_1.fastq.gz,https://ftp.sra.ebi.ac.uk/vol1/fastq/ERR101/054/ERR10181254/ERR10181254_2.fastq.gz,auto

tsvCopy code
sample          replicate    species               sex  age                            genotype              tissue        ecotype
SAMEA111350649  ERR10181254  Arabidopsis thaliana  NA   mature non-fertilized ovule    wild type genotype    plant ovule   Col-0

Output Structure

Copy code
output_directory/
├── report.md              # Commands run and what each returned
├── result.json            # Machine-readable envelope
├── samplesheet.csv        # Pipeline-ready nf-core samplesheet
├── fastq_md5.tsv          # (optional) archive MD5 per URL; URL sheets only
├── download_ena.sh        # Runnable bash + SLURM download script
├── tables/
│   └── metadata.tsv       # Standardised sample x run table
├── downloads/             # (optional) only with --command download
└── reproducibility/
    ├── commands.sh        # Exact command to reproduce
    ├── environment.yml    # Environment snapshot
    └── checksums.sha256   # SHA-256 of every artifact

Dependencies

Required: Python >= 3.10 only. The vendored client is standard library (urllib, csv, re, html) — nothing to pip install.

Optional: wget or curl on the machine that runs the generated download script; sbatch if it is submitted rather than run.

Gotchas

• Gotcha 1: You will want to trust the _1/_2 filename heuristic for a 10x run. Do not. When the technical reads are separate files the heuristic silently picks the barcode read and drops the cDNA read, and the pipeline then quantifies nothing. Pass --read-map explicitly: 3 files (single index) → 2,3; 4 files (dual index) → 3,4.

• Gotcha 2: You will read the age column as an age. It is the mapped target for developmental stage too, so for the demo study it holds "mature non-fertilized ovule". The core columns are a harmonisation, not a schema the source guarantees — check the promoted extra columns before drawing conclusions.

• Gotcha 3: --read-map validates positions 1–9, not 1–4. A typo like 1,9 passes validation and fails much later with a missing-file error. Check the run's actual file count in runs output first.

• Gotcha 4: An empty result is normal, not an error. Controlled-access studies and runs not mirrored to ENA return no fastq_ftp, and the skill says so rather than inventing links. Those runs are still reachable from SRA with sra-tools (prefetch + fasterq-dump), which is an external tool, not a ClawBio skill — do not promise a skill that does not exist.

• Gotcha 5: download-script writes a script and downloads nothing. Never run or submit it without telling the user the file count and total size first, and never treat one approval as covering a later run. The script checks every file against ENA's fastq_md5 (from fastq_md5.tsv, written by samplesheet) and stops on a mismatch. You will want to re-run it to "resume" past that error. Do not. A resumed transfer of a corrupt file is still corrupt: delete the named file first.

• Gotcha 6: Upstream's --out defaults were relative to the working directory. Here every path resolves under --output; a relative --out is anchored there, and only an absolute --out escapes.

• Gotcha 7: If metadata/runs work but download hangs or fails with a connection timeout, suspect a firewall, not a bug — but check the [download] line first. download prints the file count and total size before it blocks, then produces no further output until it finishes — a 709 GB study is not hung, it is 709 GB. A firewall gives a connection timeout; a slow link gives silence. Only the timeout is the case described here. Metadata comes from www.ebi.ac.uk (Portal + Browser); FASTQ bytes come from ftp.sra.ebi.ac.uk. Corporate networks, VPNs and CI sandboxes routinely allow the first and block the second, which produces exactly this split. Both hosts must be allowlisted on TCP/443 — this skill never speaks the FTP protocol despite the hostname, so opening FTP ports achieves nothing. See docs/data-handling.md. Confirm with curl -sI https://ftp.sra.ebi.ac.uk/vol1/ -o /dev/null -w '%{http_code}\n': 200 means reachable, 000 means blocked. The same applies to a generated download-script run on a compute node, which often has stricter egress than the login node it was written on.

• Gotcha 8: --limit means different things per command, so it has no global default. On search it caps hits (20 when omitted). On report it caps data rows, and omitting it means all rows — passing a default here once truncated a 95-run study to 20 with status: ok, which is indistinguishable from a complete report. runs, metadata-table and samplesheet always fetch everything. If a report comes back with exactly --limit rows you get a truncation warning, in both the terminal and report.md; re-run with --limit 0 to be certain.

Safety

• Local-first: no user genetic data is ever transmitted. This skill sends a public accession or the query you typed to www.ebi.ac.uk and receives public archive data. Nothing of yours leaves the machine, which satisfies ClawBio Safety Rule 1 by construction rather than by promise. See docs/data-handling.md.
• Credentials: none. ENA's public Portal needs no key.
• Execution is opt-in: download-script only writes a file. --run and --submit are off by default and must be confirmed by the user each time.
• Disclaimer: every report carries the ClawBio medical disclaimer.
• Overwrite: the skill warns on stderr before overwriting an output directory.
• Audit trail: every run writes reproducibility/.

Agent Boundary

The agent dispatches, explains, and asks before anything is executed. The skill executes. The agent must not invent accessions or FASTQ URLs, must not guess a --read-map it has not verified against the run's file count, and must not run or submit a generated script without explicit confirmation.

Integration with Bio Orchestrator

Trigger conditions: the orchestrator routes here on an ENA accession (PRJEB, ERR, ERX, SAMEA, ERZ) or an explicit mention of ENA.

Chaining partners:

• sra-tools (external, not a ClawBio skill): the route for 10x/Chromium reads, where only fasterq-dump exposes the read structure faithfully, and for runs not mirrored to ENA.
• geo-fetch: GEO resolves to ENA for its FASTQ links, so the two agree.
• nfcore-rnaseq-wrapper / nfcore-scrnaseq-wrapper: the natural consumers of the samplesheet.csv this skill writes.
• article-data-fetcher: upstream producer. It resolves a DOI or PMID to the repository accessions a paper deposited, ENA among them. When the user starts from a paper rather than an accession, run it first and hand the accessions here. It downloads files and writes a manifest.json, but it does not harmonise sample annotation into metadata.tsv or emit a pipeline-ready samplesheet.csv — that is this skill's job, so the two chain rather than compete.

Maintenance

• Update trigger: when an archive host changes an endpoint this skill calls. Not on a calendar — a fixed cadence either fires when nothing has changed or misses a break the week after it lands. The staleness signals below are the trigger.
• How a break surfaces: from a live call, not from CI. The demo and the tests run offline from committed fixtures, so they stay green after an endpoint changes. Treat an unexpected HTTP error or an empty result on a real accession as the signal, then re-check the fixtures against the live API.
• Staleness signals: DEFAULT_RUN_FIELDS names a field the Portal drops; the Browser XML schema changing SAMPLE_ATTRIBUTES; nf-core changing its samplesheet column contract.
• Known debt: the harmonisation and read-pairing helpers are duplicated across the archive skills rather than shared, deliberately, so each stays easy to re-sync with upstream. Factor them out only if upstream does.
• Deprecation: if ENA ships an official Python client covering these commands, wrap it instead of the REST API.

Citations

• ENA Portal API and ENA Browser API; the endpoints used.
• nf-core/rnaseq and nf-core/scrnaseq; the samplesheet contracts.
• Demo fixture: PRJEB56029, Arabidopsis thaliana RNA-seq; ENA records are open data.
• Ported from UKDRI/informatics_data_skills @ 7cc3e6e (ena/, plus fastq-download-script/ folded in as download-script), © 2026 UK Dementia Research Institute, MIT.