AutoGineer
Generated banner for Human Protein Atlas Database
Generated logo for Human Protein Atlas Database

Human Protein Atlas Database

Google DeepMindOfficialPublished by Google DeepMind

  • Skill
  • Tool integration
  • Science and research
  • Apache-2.0
  • Low risk
  • Has scripts
Download ZIP

From the author

Use when you want to retrieve semi-quantitative protein expression and spatial localisation data from the Human Protein Atlas (HPA).
— the description by Google DeepMind

AutoGineer summary

Written by AutoGineer from the skill’s own files, not by the author

The Human Protein Atlas (HPA) database integration allows researchers to access semi-quantitative protein expression data derived from immunohistochemistry (IHC) across normal human tissues and cancer types. This tool bridges the gap between transcriptomic data and actual protein presence, providing insights into subcellular localization and systemic protein abundance. It requires the use of a provided CLI wrapper to ensure fair usage and proper data handling.

What it does

  • Maps standard gene symbols to Ensembl IDs for database queries.
  • Retrieves semi-quantitative protein expression levels in specific human tissues.
  • Identifies subcellular organelles and structures where a protein is localized.
  • Fetches comprehensive metadata entries for specific genes.
  • Performs advanced searches based on expression criteria and specificity categories.

When to use it

  • When you need to confirm if a protein product exists for a specific gene.
  • When investigating the subcellular localization of a protein within a cell.
  • When comparing RNA-seq consensus data with actual protein expression levels.
  • When filtering for genes based on tissue-specific or brain-region-specific expression patterns.

Advantages

  • Provides protein-level validation beyond simple mRNA transcription data.
  • Standardizes queries through a robust CLI wrapper.
  • Supports complex filtering via the HPA search query syntax.
  • Automates the retrieval of large-scale biological datasets into JSON format.

Usage examples

“Get the Ensembl ID for the TP53 gene.”
The agent executes the resolve-ensembl-id command to map the symbol to its corresponding Ensembl ID, saving the result to a JSON file.
“Where is the protein encoded by ENSG00000141736 located within the cell?”
The agent uses the get-subcellular-location command to retrieve and report the specific organelles or structures associated with the protein.
“Find genes that are enriched in the amygdala.”
The agent uses the search-hpa command with the query 'brain_category_rna:amygdala' to return a list of matching genes.

Install

  1. Download the skill as a ZIP:

    https://autogineer.com/agent-skills/google-deepmind/human-protein-atlas-database/download

  2. Unzip it into the cross-client skills folder of your project:

    mkdir -p .agents/skills && unzip human-protein-atlas-database.zip -d .agents/skills

Documentation by Google DeepMind

Human Protein Atlas (HPA) Database Integration

This skill provides semi-quantitative protein expression and spatial localisation data from the Human Protein Atlas (HPA). While RNA-seq (e.g., GTEx) tells us if a gene is being transcribed, HPA confirms if the protein product actually exists, where it is located within the cell (e.g. nucleus vs cytoplasm), and its concentration in systemic blood circulation. The data is based on Immunohistochemistry (IHC) across normal human tissues and cancer types.

Prerequisites

  1. uv: Read the uv skill and follow its Setup instructions to ensure uv is installed and on PATH.

  2. User Notification: If .licenses/human_protein_atlas_database_LICENSE.txt does not already exist in the workspace root directory then (1) prominently notify the user to check the terms at https://www.proteinatlas.org/about/licence, then (2) create the file recording the notification text and timestamp.

When to Use

Use this skill when you need to:

  • Map a gene symbol to its Ensembl ID for HPA queries.

  • Retrieve the semi-quantitative protein abundance in normal human tissues and cancer types based on IHC staining (High, Medium, Low, or Not Detected).

  • Find the specific organelles or subcellular structures where a protein has been localized (e.g., nucleoplasm, mitochondria).

  • Check the consistency/agreement between RNA-seq consensus and protein expression levels.

  • Search for genes based on specific protein expression criteria (e.g., "elevated in amygdala" or "secreted proteins").

Do NOT use when you need to:

  • Query eQTLs, pQTLs, or any variant-level associations. HPA provides wild-type expression data and knows nothing about QTLs.

  • Query gene expression in non-human species. HPA is strictly for human proteins.

  • Retrieve purely quantitative RNA expression without interest in the protein product (consider using the GTEx skill instead).

Command Selection Guide

Pick the right command on the first try. Match the user's input to the correct subcommand below.

  • Map a gene symbol to Ensembl ID: resolve-ensembl-id

  • Get tissue protein expression levels: get-tissue-expression

  • Get subcellular location of a protein: get-subcellular-location

  • Get the full HPA metadata entry for a gene: get-atlas-entry

  • Search HPA for genes matching specific criteria: search-hpa

Quick Start

# Map the ERBB2 gene symbol to its Ensembl ID
uv run scripts/hpa_cli.py resolve-ensembl-id ERBB2 --output /tmp/erbb2_id.json

# Get subcellular location by Ensembl ID
uv run scripts/hpa_cli.py get-subcellular-location ENSG00000141736 --output /tmp/erbb2_location.json

All subcommands write JSON to disk. Always save output in the /tmp/ directory. The default output file is /tmp/hpa_output.json if --output is not specified.

Commands

1. resolve-ensembl-id — Gene Symbol → Ensembl ID

Maps a common gene symbol (e.g., "TP53", "ERBB2") to its Ensembl gene ID. HPA endpoints are strictly Ensembl-based.

uv run scripts/hpa_cli.py resolve-ensembl-id TP53 --output /tmp/tp53_id.json

Arguments:

  • gene_symbol (positional): The standard gene symbol (e.g., "TP53").

  • --output: Output file path (default: /tmp/hpa_output.json).

2. get-tissue-expression — Get Tissue Protein Levels

Returns a list of tissues and their corresponding protein expression levels (High, Medium, Low, or Not Detected) based on IHC staining.

uv run scripts/hpa_cli.py get-tissue-expression ENSG00000130234 \
  --tissues "duodenum,thyroid gland" --output /tmp/tissue_expr.json

Arguments:

  • ensembl_id (positional): The Ensembl Gene ID.

  • --tissues: Comma-separated list of tissues to filter by (optional, defaults to all available tissues).

  • --output: Output file path (default: /tmp/hpa_output.json).

3. get-subcellular-location — Get Subcellular Location

Retrieves the specific organelles or cellular structures where the protein has been localized.

uv run scripts/hpa_cli.py get-subcellular-location ENSG00000141736 \
  --output /tmp/subcellular.json

Arguments:

  • ensembl_id (positional): The Ensembl Gene ID.

  • --output: Output file path.

4. get-atlas-entry — Get Full HPA Entry

Fetches the full metadata for a gene, including IHC scores, RNA-seq consensus, and subcellular location.

uv run scripts/hpa_cli.py get-atlas-entry ENSG00000254647 \
  --output /tmp/ins_entry.json

Arguments:

  • ensembl_id (positional): The Ensembl Gene ID.

  • --format: Format of the returned entry, e.g., json (default: json).

  • --output: Output file path.

5. search-hpa — Search by Attribute

Allows filtering for genes based on specific criteria (e.g., "elevated in amygdala").

uv run scripts/hpa_cli.py search-hpa \
  --query "brain_category_rna:amygdala" \
  --output /tmp/search_results.json

Arguments:

  • --query: The search query string. Refer to references/search-api.md for details.

  • --output: Output file path.

Core Rules

  • Use the Wrapper: ALWAYS execute the provided helper scripts to query the database rather than accessing the database directly. The scripts automatically enforce fair use and implement retry logic.

  • Notification: If this skill is used, ensure this is mentioned in the output.

API Versioning

The HPA website at www.proteinatlas.org always serves the latest data release. Older archived versions can be accessed via vNN.proteinatlas.org (e.g., v24.proteinatlas.org), while the current version's subdomain redirects to www.proteinatlas.org. This skill's scripts query the latest version by default.

Common Errors

  • If no results are returned, confirm the query is detailed enough starting with the api reference in references/search-api.md

  • If you cannot find the results, search the web for example HPA queries and use these to construct a better query.

  • The output is usually large. Use jq or write your own python data parsing library to process the search results. Never output to stdout, or cat the output file.

Files

references/search-api.md

Human Protein Atlas Search Query API Reference

This document provides a comprehensive guide to constructing search queries for the Human Protein Atlas (HPA). The search engine allows users to filter the HPA database based on protein expression, mRNA levels, subcellular localization, and functional classifications.

1. Core Syntax Overview

The HPA search engine follows a standard key-value pair format. All queries are case-insensitive.

IMPORTANT: Terms with spaces do not need to be enclosed in double quotes (e.g. protein_class:Transcription factors), and adding quotes may actually break the query. When a field has sub-categories, values for them are separated by a semi-colon ;. Multiple selections within a sub-category are separated by a comma ,.

  • Field Search: field:value

    • Example: chromosome:12

  • Field with Subfields: field:subvalue1;subvalue2

    • Example: tissue_category_rna:Brain;Tissue enriched

  • Multiple selections: field:subval1;sel1,sel2

    • Example: tissue_category_rna:Any;Tissue enriched,Group enriched

  • Boolean AND: term1 AND term2

    • Example: protein_class:Enzymes AND chromosome:X

  • Boolean OR: term1 OR term2

    • Example: tissue_category_rna:Any;Tissue enriched OR tissue_category_rna:Any;Tissue enhanced

  • Boolean NOT: term1 NOT term2

    • Example: protein_class:Enzymes NOT chromosome:1

  • Grouping: ( ... )

    • Example: (tissue_category_rna:Any;Tissue enriched OR tissue_category_rna:Any;Group enriched) AND chromosome:1

2. Specificity Classifications

The HPA categorizes gene expression based on RNA-seq data across tissues (General Atlas), brain regions (Brain Atlas) and single cell types (Single Cell Atlas). These categories are built out of two dropdowns in the UI: the region/tissue to filter on, and the specificity category.

RNA Tissue Specificity (tissue_category_rna)

Filters based on the general tissue distribution across the entire human body. Format: tissue_category_rna:<Tissue Name>;<Specificity Category>

  • Specificity Categories:

    • Tissue enriched: Genes with mRNA levels at least 4-fold higher in a single tissue compared to all others.

    • Group enriched: Genes with mRNA levels at least 4-fold higher in a group of 2-5 tissues compared to all others.

    • Tissue enhanced: Genes with mRNA levels at least 4-fold higher in a tissue compared to the average of all others.

    • Low tissue specificity: Low tissue specificity; detected in many tissues.

    • Not detected: Not detected in tissues.

Example: tissue_category_rna:Liver;Tissue enriched Example: tissue_category_rna:Any;Tissue enriched,Group enriched

RNA Brain Region Specificity (brain_category_rna)

Filters based on distribution across different brain structures. Format: brain_category_rna:<Brain Region>;<Specificity Category>

  • Specificity Categories:

    • Region enriched: Genes with mRNA levels at least 4-fold higher in one brain region compared to all others.

    • Group enriched: mRNA levels at least 4-fold higher in 2-5 brain regions.

    • Region enhanced: mRNA levels at least 4-fold higher in a brain region compared to the average of others.

Example: brain_category_rna:Amygdala;Region enriched,Group enriched

Pro-Tip: If you are specifically looking for proteins that are unique to the brain compared to the rest of the body, combine tissue_category_rna and brain_category_rna fields.

3. Commonly Used Query Fields

Below is a reference of the most frequently used fields for filtering the database.

  • gene_name: Search by the official HGNC symbol.

    • Examples: gene_name:APOE, gene_name:TP53

  • chromosome: Filter by the genomic location.

    • Examples: 1, 2, ..., X, Y, MT

  • protein_class: Functional classification of the protein.

    • Examples: Enzymes, Transcription factors, FDA approved drug targets

  • subcell_location: Main localization based on ICC staining.

    • Examples: Nucleoplasm, Mitochondria, Cytosol

  • predicted_location: Filter for secreted or intracellular logic.

    • Examples: Predicted secreted proteins, Predicted membrane proteins

  • cancer_category_rna: Expression in specific cancers.

    • Example: cancer_category_rna:Breast cancer;Cancer enriched

  • ihc_ab_validation: The validation level of the IHC antibody data.

    • Examples: Supported, Approved, Enhanced - Independent

4. Constructing Complex Queries

Scenario A: Finding Specific Brain Markers

To find genes that are Enriched in a specific brain region (e.g., Hypothalamus) and localized to the Vesicles (for secretory/synaptic paths): brain_category_rna:Hypothalamus;Region enriched AND subcell_location:Vesicles

Scenario B: Finding Elevated Genes in a Region

To find genes that show an elevated protein expression level in the amygdala compared to other regions of the human brain:

brain_category_rna:Amygdala;Region enriched,Group enriched,Region enhanced
AND sort_by:Tissue specific score
Scenario C: Filtering for Validated Enzymes on Chromosome 1

To find Enzymes with Approved reliability (high-quality IHC data) located on Chromosome 1: protein_class:Enzymes AND chromosome:1 AND ihc_ab_validation:Approved

Scenario D: Identifying Tissue-Specific Transcription Factors

To find Transcription factors that are either Tissue Enriched or Group Enriched across any tissue:

protein_class:Transcription factors
AND tissue_category_rna:Any;Tissue enriched,Group enriched

Note on Data Versions: The HPA is updated periodically. The specificity categories remain consistent, but the underlying RNA-seq datasets (e.g., HPA vs. GTEx) may yield slightly different results if you specify the data source.

Requirements and compatibility

Works with: Claude Code, Claude apps, OpenAI Codex, GitHub Copilot, Cursor, Gemini CLI, Kiro, OpenCode, Goose, Amp, Roo Code, JetBrains Junie, Factory Droid, Devin (ex-Windsurf), Mistral Vibe, OpenHands, Letta

“Works with” means the tool reads the open Agent Skills format; a skill written for one tool may use features others lack.

Needs: Needs Python

Security

What our scanner noticed· Low risk

  • Scripts that your agent may run

This skill includes scripts that your agent may run. Read them before you use it.

A static scan is not a guarantee. Skills are third-party content: review a skill before you let an agent use it.

Attribution and licence

Published by
Google DeepMind
Publisher
Google DeepMind (Official)
Source
google-deepmind/science-skills · skills/human_protein_atlas_database · commit 4de6786
Licence
Apache License 2.0 (redistribution allowed)
  • LICENSE (added by AutoGineer from LICENSE)
  • ATTRIBUTION.md (written by AutoGineer)
Synced

AutoGineer is not affiliated with or endorsed by Google DeepMind or Google DeepMind.

Download ZIP

The ZIP holds the skill’s files, its licence files and an ATTRIBUTION.md. Download human-protein-atlas-database.zip