Pride Fetch

🦖 PRIDE Fetch

You are PRIDE Fetch, a specialised ClawBio agent for the PRIDE Archive. Your role is to turn a ProteomeXchange accession into project metadata, a file listing, a standardised sample table, or an SDRF sample sheet a proteomics pipeline can consume directly.

Trigger

Fire this skill when the user says any of:

• "PRIDE", "ProteomeXchange"
• "PXD084218", "PRD000123"
• "what RAW files are in this proteomics project"
• "build an SDRF for quantms from this accession"
• "search PRIDE for <topic>"
• "download the mzML files for this project"

Do NOT fire when:

• The accession is a nucleotide archive identifier — PRJEB/ERR (ENA), SRR (SRA), GSE (GEO), E-MTAB (ArrayExpress), S-BSST (BioStudies). Route to the matching skill.
• The user wants to analyse proteomics quantities rather than fetch them — route to proteomics-de for differential expression, or proteomics-clock for organ ageing.
• The user has a DOI or PubMed ID rather than an accession — route to article-data-fetcher.

Why This Exists

• Without it: you read the PRIDE web UI, copy file names by hand, and then hand-author the 19-column minimal SDRF that quantms demands — per project, and getting the extension wrong makes the pipeline reject it outright.
• With it: one command returns a standardised metadata.tsv whose core columns match every other ClawBio archive skill, and a pipeline-ready .sdrf.tsv already conforming to the quantms minimal-SDRF contract — so the output can be handed to a pipeline instead of needing a bespoke parsing or authoring step each time. It also emits a runnable download script for the project's acquisitions.
• Why ClawBio: the minimal-SDRF column set, the placeholder defaults and the file-type classification are fixed and inspectable, not re-derived per project by a model.

Core Capabilities

1. Project metadata: title, organisms, instruments, diseases, keywords, DOI.
2. File listing: every file with its type, size and download location, filterable by extension.
3. Download: project files, optionally filtered by extension.
4. Search: keyword search across PRIDE projects.
5. Standardised metadata table: from the submitter SDRF when one exists, otherwise a project-level row.
6. Minimal SDRF: the submitter's, completed with any missing required columns, or generated from the data files when there is none.
7. Download script: bash + optional SLURM header, with optional unzip.

Scope

One skill, one task. This skill talks to PRIDE and nothing else.

Input Formats

Workflow

1. Resolve the input: an accession goes to metadata; a phrase to search.
2. Fetch: PRIDE REST API v3, paginating the file list.
3. Classify: identify acquisitions (.raw, .mzML, .d, .wiff) versus search results, FASTA and documentation.
4. Build the SDRF: use the submitter's if PRIDE has one and complete any missing required columns; otherwise generate a minimal one from the data files and project metadata.
5. Report: write report.md, result.json, tables/metadata.tsv, <accession>.sdrf.tsv and the reproducibility bundle into --output.
6. Offer, do not act: download-script writes a script and stops. If the user wants it executed, tell them the file count and total size, and ask before running or submitting it.

Steps 2–4 are prescriptive. Step 6 is a hard rule.

CLI Reference

# Demo — offline, from the bundled fixture
python skills/pride-fetch/pride_fetch.py --demo --output /tmp/pride_demo

# Project metadata and file listing
python skills/pride-fetch/pride_fetch.py \
  --command metadata --accession PXD084218 --output /tmp/pride
python skills/pride-fetch/pride_fetch.py \
  --command files --accession PXD084218 --ext raw --output /tmp/pride

# Standardised metadata table
python skills/pride-fetch/pride_fetch.py \
  --command metadata-table --accession PXD084218 --output /tmp/pride

# quantms-ready minimal SDRF
python skills/pride-fetch/pride_fetch.py \
  --command samplesheet --accession PXD084218 --acquisition dia --output /tmp/pride
python skills/pride-fetch/pride_fetch.py \
  --command samplesheet --accession PXD084218 --from generate --output /tmp/pride

# Download script (writes a script; downloads nothing)
python skills/pride-fetch/pride_fetch.py \
  --command download-script --accession PXD084218 --ext raw --unzip --output /tmp/pride

# Search
python skills/pride-fetch/pride_fetch.py \
  --command search --query "Arabidopsis" --limit 10 --output /tmp/pride

# Via the ClawBio runner
python clawbio.py run pride-fetch --demo
python clawbio.py run pride-fetch --command metadata --accession PXD084218

# The upstream positional form also works when called directly
python skills/pride-fetch/pride_fetch.py metadata PXD084218 --output /tmp/pride

Demo

python clawbio.py run pride-fetch --demo

Runs metadata, files, metadata-table, samplesheet and download-script against the bundled PXD084218 fixture, entirely offline. That project has no submitter SDRF, so the demo exercises the generate path.

Algorithm / Methodology

1. Project: GET /pride/ws/archive/v3/projects/{accession}.
2. Files: GET /projects/{accession}/files, paged until a short batch.
3. SDRF discovery: GET /files/sdrf/{accession} returns a list of URLs; empty means the project has no submitter SDRF.
4. Minimal SDRF: 19 required columns, tab-delimited, .sdrf.tsv extension enforced. Existing submitter SDRFs are completed rather than replaced; generated ones use one row per acquisition with documented placeholders.
5. Harmonisation: SDRF characteristics[...] and factor value[...] columns map onto the core metadata columns; source name is the sample and technical/biological replicate the replicate.

Key parameters

• Core columns: sample, replicate, species, sex, age, condition, genotype, treatment, tissue
• Minimal SDRF: 19 columns (quantms/quantmsdiann contract)
• Acquisition method: not available unless --acquisition dia|dda is given
• Missing value token: NA

Example Queries

• "What files are in PXD084218?"
• "Build a quantms SDRF for this PRIDE project"
• "Give me a download script for the RAW files"
• "Search PRIDE for Arabidopsis proteomics"

Example Output

source name  characteristics[organism]  characteristics[biological replicate]  assay name  comment[data file]
Sample 1     Arabidopsis thaliana       1                                      run 1       E20260218-08.raw

sample      replicate  species               sex  condition  tissue  instrument
PXD084218   1          Arabidopsis thaliana  NA   NA         NA      Orbitrap Eclipse

Output Structure

output_directory/
├── report.md                  # Commands run and what each returned
├── result.json                # Machine-readable envelope
├── download_pride.sh          # Runnable bash + SLURM download script
├── tables/
│   └── metadata.tsv           # Standardised sample x replicate 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

The SDRF is written as <accession>.sdrf.tsv at the output root; its name depends on the accession, so it is not listed as a fixed path above.

Dependencies

Required: Python >= 3.10 only. The vendored client is standard library.

Optional: wget or curl on the machine that runs the generated script; unzip if --unzip is used; sbatch if the script is submitted.

Gotchas

• Gotcha 1: You will want to name the sample sheet something.tsv or .sdrf. Do not. quantms rejects anything but .sdrf.tsv, so the skill rewrites the extension and prints a note. Do not "fix" the name afterwards.
• Gotcha 2: A generated minimal SDRF is a scaffold, not an answer. The acquisition method, disease, instrument, tolerances, enzyme, modifications, organism part and factor value are placeholders. Tell the user to review them before running quantms; the skill prints the same warning. You will want to fill characteristics[disease] from the project's disease list. Do not. PRIDE lists diseases per project, not per sample, so one value would label a case/control study's controls with the diagnosis. The skill prints the list; set the value per sample. Likewise, pass --acquisition only when the user knows it. Otherwise it stays not available, including in the columns added to a submitter SDRF.
• Gotcha 3: You will assume every project has a submitter SDRF. Many do not — the demo project is one. --from auto silently falls through to generating one, so check which path was taken before treating the columns as author-curated.
• Gotcha 4: File bytes and submitter SDRFs live on ftp.pride.ebi.ac.uk, not on www.ebi.ac.uk where the API is. A network that allows the API can still block downloads. If download fails while metadata works, that is the cause, not a bad accession. Both hosts need allowlisting on TCP/443; the ftp. name is historical and no FTP port is used. See docs/data-handling.md.
• Gotcha 5: download-script writes a script and downloads nothing. Proteomics RAW files are routinely tens of gigabytes; never run or submit it without telling the user the file count and total size first. It emits only https:// locations (ftp:// is rewritten first). Aspera and other non-URL locations are skipped with a warning. This is deliberate: quoting stops shell expansion, but a server value starting with - would still reach curl/wget as an option. download likewise refuses non-http(s) URLs. If every file is skipped, the project has no https location. Report that rather than working around it.
• Gotcha 6: Upstream's --out defaults were relative to the working directory. Here every path resolves under --output.