marker-gene-validator

name	marker_gene_validator
description	Validate whether a marker set supports a proposed cell type or cell state and explain the main caveats.
category	bio/single_cell_rna
version	1
requires_tools	["search_knowledge_base","ncbi_eutils","uniprot_api","python_repl"]
requires_network	true
user_invocable	true
tags	["markers","validation","annotation","cell-state"]
aliases	["marker_panel_validator"]
species	any
modality	single_cell_rna
stage	annotation
stability	stable
safety_level	low

Marker Gene Validator

Purpose

Check whether a proposed marker set really supports the claimed cell identity or state.

When to use

Use this skill when the user already has a marker panel or top DE genes and wants to know whether the label is convincing.

Required inputs

• markers: a list of marker genes
• proposed label: cell type or state
• context (optional): species, tissue, disease, stimulation, or assay context

Steps

1. Restate the proposed label, species, tissue, and disease or stimulation context because the same markers can mean different things across systems.
2. Search local guidance such as marker panel notes with search_knowledge_base before leaning on generic marker heuristics.
3. Use ncbi_eutils and uniprot_api to confirm unclear, surprising, or multifunctional markers when necessary.
4. Use python_repl to organize markers into strongly supportive, weakly supportive, and conflicting groups if the list is long.
5. Return a confidence judgment that clearly separates well-supported markers from context-dependent or contradictory ones.

Output format

• Biological context or assumptions: proposed label, species, tissue, and any inferred activation or disease context.
• Evidence or source basis: which search_knowledge_base, ncbi_eutils, uniprot_api, and python_repl checks support the call.
• Marker assessment: supportive markers, conflicting markers, and confidence in the proposed label.
• Caveats or ambiguity: lineage overlap, activation-state confounding, or insufficient marker coverage.
• Recommended next step: what marker, modality, or follow-up validation would most reduce uncertainty.

Failure modes

• Too few markers: say confidence is limited.
• Conflicting markers: return alternative labels or mixed-state possibilities.
• Broad label: ask for more context if the claim is underspecified.

Examples

• "Do PDCD1, TOX, TIGIT, LAG3 support an exhausted CD8 T-cell label?"
• "Are these markers enough to call this cluster macrophages?"