Research protein

Skill mriffle/findings-ai-collab-workflow/skills/research-protein

How to research a protein or gene for a Findings Workflow project using the authoritative databases (UniProt, PDB/AlphaFold, STRING, GO), with identifier and fact verification. Use when a finding involves a specific protein/gene and needs grounded background. Output feeds a research-finding.From its SKILL.md

Install

npx -y skills add mriffle/findings-ai-collab-workflow --skill research-protein

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

2.2 KB, as published. Nobody here has run it

Researching a protein / gene

Goal: grounded, normalized, verifiable background on a specific protein or gene. Every claim is sourced; every identifier is canonical and confirmed.

Authoritative sources

UniProt — the canonical protein record: accession, recommended name, gene name, function, subcellular location, PTMs, sequence features. Use the UniProt accession as the canonical id (db: uniprot).
HGNC — the canonical human gene symbol (db: hgnc). Resolve symbol↔accession explicitly; do not assume.
PDB / AlphaFold — experimental and predicted structures (cite the PDB id / AlphaFold entry).
STRING — functional interaction partners and evidence channels (treat as evidence-weighted, not ground truth).
GO — function/process/component annotations, with evidence codes (db: go); note the evidence code, since IEA (electronic) ≠ experimental.
Reactome — pathway membership (db: reactome).

Procedure

Pin the identity first. Confirm the exact UniProt accession and gene symbol for the right organism/isoform. Identifier integrity matters — the wrong accession poisons everything downstream.
Gather function/role from UniProt + GO + Reactome, recording the source for each claim and the GO evidence code where relevant.
Interactions/structure from STRING / PDB / AlphaFold as the question needs, noting confidence/evidence.
Verify each claim against its source before recording it. Distinguish experimentally supported from predicted/electronic annotation.

Output

A research-finding (templates/research-finding.md) with the entity normalized to canonical IDs (verified), each claim sourced (database record or paper), and a non-empty references list. It goes to the research-reviewer, which re-checks the IDs and every reference. Never attach an unverifiable accession or citation — record the claim as unverified instead.