version: "1.0.0" name: normal-tissue-expression-for-gene description: Retrieve and summarize normal tissue expression for a single human gene using GTEx, including full tissue-level expression, publication-quality barplot output, and a simple expression-pattern classification such as universally expressed, tissue-specific, mixed, or non-expressed.

Normal Tissue Expression for Gene

Purpose

Use this skill to retrieve and summarize normal tissue expression for a single gene.

This skill is intended for common daily bioinformatics questions such as:

• In which normal tissues is this gene highly expressed?
• Is this gene broadly expressed across tissues or tissue-specific?
• Is this gene lowly expressed or effectively non-expressed in GTEx?
• What does the expression distribution across all normal tissues look like?

This skill is single-gene focused by design. It is optimized for fast, interpretable inspection of one gene at a time rather than large matrix-style gene-list analysis.

Scope

This skill covers:

• normal tissue expression
• GTEx-based tissue-level summaries
• publication-quality visualization
• basic expression-pattern classification

This skill does not cover:

• cancer cell line expression
• tumor expression
• protein abundance
• gene essentiality
• pathway enrichment
• multi-gene heatmaps as the primary use case

Those should be handled by separate skills.

Data source

Primary data source:

• GTEx normal tissue expression

The implementation should prefer official GTEx API access when available.

Recommended retrieval flow:

1. Resolve the queried gene against GTEx gene reference data
2. Retrieve median tissue expression values for that gene
3. Normalize the returned fields into a consistent table
4. Generate summary outputs and figures

When to use

Use this skill when the user asks things like:

• “Show me normal tissue expression of TP53.”
• “Which tissues highly express EGFR?”
• “Is GATA3 tissue-specific?”
• “How broadly is MYC expressed across tissues?”
• “Plot GTEx tissue expression for this gene.”

When not to use

Do not use this skill when the user asks for:

• cancer cell line expression
• tumor expression
• gene-list heatmaps as the main deliverable
• protein expression
• CRISPR dependency
• gene annotation unrelated to tissue expression

Use another skill for those requests.

Input

Required input

--gene

A single gene symbol.

Examples:

• TP53
• EGFR
• ERBB2
• GATA3

Optional input

--top-n

Number of top tissues to summarize in the text summary.

Default:

• 10

--expr-detect-threshold

Expression threshold used to define whether a tissue is considered “detected” or expressed.

Default:

• 1.0

This threshold is also shown on the barplot.

--outdir

Output directory for all generated files.

Why this skill is gene-level instead of gene-list-level

This skill is intentionally centered on one gene at a time, because the most common practical use case is:

1. the user has a gene of interest,
2. they want to quickly inspect its tissue distribution,
3. they want an interpretable classification and figure.

For this use case, a single-gene workflow is usually more useful than a matrix-first workflow.

A multi-gene normal-tissue-expression skill can be built separately if needed.

Outputs

A successful run should produce:

• one tissue-level expression table
• one concise summary text file
• one summary TSV file
• one publication-quality barplot in PNG
• one publication-quality barplot in PDF

Typical output files

• <GENE>.gtex_tissues.tsv
• summary.txt
• gene_expression_summary.tsv
• <GENE>.gtex_all_tissues.png
• <GENE>.gtex_all_tissues.pdf

Output table structure

Tissue-level table

The per-gene tissue table should contain:

• gene_symbol
• gencode_id
• tissue
• expression
• rank_within_gene

Summary table

The summary TSV should contain:

• query_gene
• resolved_gene_symbol
• gencode_id
• expression_category
• max_expression
• median_expression_across_tissues
• num_tissues
• num_detected_tissues
• fraction_detected_tissues

Core workflow

Step 1. Parse and normalize the input gene

• Read the gene from --gene
• Normalize formatting
• Treat the query as a human gene symbol

Step 2. Resolve the gene in GTEx

• Query GTEx reference gene information
• Resolve:
• canonical gene symbol
• GENCODE ID

If the gene cannot be resolved, fail clearly.

Step 3. Retrieve GTEx normal tissue expression

• Query GTEx median tissue expression
• Normalize the response into a consistent table with:
• tissue
• expression

Step 4. Rank and summarize tissues

• Sort tissues by expression
• Keep all tissues in the output table
• Use top-n only for the short text summary

Step 5. Classify the expression pattern

Assign one of the following categories:

• universally-expressed
• tissue-specific
• mixed
• non-expressed
• no-data

Step 6. Generate a publication-quality barplot

• Show all tissues by default
• Use a horizontal barplot
• Draw a threshold line for the expression detection threshold
• Save as both PNG and PDF

Step 7. Write output files

Write the tissue table, summary files, and figure to outdir.

Expression classification logic

This skill should provide a simple, interpretable classification rather than an overly complicated tissue-specificity score.

Recommended logic:

non-expressed

Use when:

• the maximum tissue expression is below the detection threshold

tissue-specific

Use when:

• maximum expression is clearly high
• expression in the top tissue is much higher than the median across tissues
• only a relatively small fraction of tissues are above the detection threshold

universally-expressed

Use when:

• most tissues are above the detection threshold
• median expression across tissues is not low

mixed

Use when:

• the gene is expressed in some tissues
• but does not strongly fit either universal or highly tissue-specific behavior

no-data

Use when:

• no GTEx result is returned for the gene

Threshold guidance

The default expression detection threshold is:

• 1.0

This is a practical and commonly used threshold for judging whether expression is present at a meaningful level.

Interpretation:

• below threshold: low or not detected
• above threshold: detectable expression

The threshold should also be displayed visually on the barplot.

Plotting requirements

The barplot should be publication quality.

Plot style requirements

• horizontal barplot
• all tissues shown by default
• tissues sorted by expression
• high resolution
• readable tissue labels
• clean axes
• minimal unnecessary styling
• threshold line clearly visible

Because the plot is horizontal

Expression is on the x-axis, so the threshold should be plotted as a vertical line, not a horizontal line.

Recommended outputs

• PNG for quick viewing
• PDF for publication or slides

Execution

Entry point

python scripts/normal_tissue_expression_for_gene.py --gene "<GENE>" --outdir <output_dir>