sdrf-metascreen

SDRF Meta-Analysis Screening Protocol

You are screening datasets before annotation or autoresearch. Your job is to resolve a candidate study set, read repository metadata and the associated publication for each candidate, apply the user's inclusion criteria, extract requested study-level metadata, and write one TSV row per accession.

Do not guess. Use the MCP tools to verify every field.

This skill is complementary to sdrf:autoresearch: run it first when the user needs a more precise, publication-aware study screen before deciding which datasets should enter annotation. This skill produces a curation TSV; it does not create, validate, fix, or improve SDRF files.

The criteria and the extract fields come from the user, not from this file. Nothing here is specific to any one research question — the PRIDE fields named in Step 2 are a worked example of "pre-filter on whatever structured fields the criteria actually constrain", not a fixed list.

Step 1: Parse the Request

Normalize into these fields:

target

What dataset set to screen. Accepted forms:

• accessions:PXD001234,MSV000078958 — use the comma- or whitespace-separated accession list directly. Do not put file paths after accessions:. Both PXD… and MSV… are supported; get_project_details routes to the right repository on its own, so no per-accession branching is needed.

• Any file path (.txt, .tsv, .csv) — resolve automatically by extension:

  • .txt — one accession per line, strip whitespace, skip blank lines
  • .tsv / .csv — read the first available accession column from: id, accession, project_accession, project, or the first column

• all <category> datasets — use search_projects to discover matching datasets. Examples:

  • all PRIDE human gut metaproteomics datasets
  • all crosslinking datasets
  • all human plasma proteomics datasets

  search_projects covers PRIDE and MassIVE together by default. A repository name inside the category phrase ("all PRIDE …") is ordinary habit, not a scoping instruction — search both. Pass repository="pride" or "massive" only when the user is unambiguously scoping to one, e.g. "PRIDE only", "just MassIVE datasets", "skip MassIVE" — phrasing that names the repository as an exclusion, not just in the category sentence.