gwas-catalog-mcp

GWAS Catalog MCP Server

Overview

This MCP server provides a programmatic interface to the GWAS Catalog REST API, enabling access to GWAS study, variant, trait, and association data. The server handles large result sets automatically by providing both in-memory results and file-based storage options.

Status

🚧 Under Active Development 🚧

This project is currently under active development. Features and APIs may change without notice.

Dependencies

• uv
• mcp[cli]
• fastmcp
• requests

Directory Structure

.
├── server.py             # Main FastMCP server entrypoint
├── utils.py              # Utility functions
├── pyproject.toml        # Project metadata and dependencies
├── README.md             # Usage and documentation
├── tests/                # Test suite and test data
│   ├── run_tests.py
│   ├── input/
│   └── output/
│       ├── success/
│       └── error/
└── ...

Setup and Running

Install dependencies

uv sync

Activate the virtual environment

. .venv/bin/activate

Run the MCP server

uv run server.py

Run tests

python tests/run_tests.py

MCP Tool Specification

Tool name

• GWAS_catalog

Common Parameters

Most tools support the following common parameters:

Tool Endpoints and Parameters

Get study

Get association

Get variant

Get trait

Search variants in region

Get variants from EFO IDs

Trait variant ranking

Get study associations

Get trait studies

Get trait associations

Get associations from variant (uses GWAS Catalog REST API)

Note: This method returns all associations for a variant, including an is_gwas_significant flag indicating if the p-value meets the genome-wide significance threshold (p ≤ 5e-8). Basically, the MCP server will only return if is_gwas_significant is True.

Get region-trait associations (uses GWAS Summary Statistics API)

Note: Endpoints marked as "uses GWAS Summary Statistics API" access https://www.ebi.ac.uk/gwas/summary-statistics/api instead of the main REST API.

Output Format

All API responses follow a consistent structure:

{
  "request_url": "https://www.ebi.ac.uk/gwas/rest/api/...",
  "items": [...],  // List of results, limited by max_items_in_memory
  "total_items_aft_process": 123,  // Total number of results after processing
  "is_complete": true,  // Whether all results are included in items
  "metadata": {
    "subset_size": 100,  // Number of items in the current response (after using max_items_in_memory parameter)
    "max_items_in_memory": 5000,  // Current memory threshold
    "total_items": 150,  // Total number of items before processing
    "significant_items": 80  // Number of genome-wide significant items (if applicable)
  }
}

Large Result Sets

When results exceed max_items_in_memory:

1. A subset of results is returned in the items field
2. is_complete will be False
3. The complete dataset is automatically saved to a file
4. The response includes an output_file field with the file path

Example large result response:

{
  "request_url": "...",
  "items": [...],  // First max_items_in_memory results
  "total_items_aft_process": 10000,
  "is_complete": false,
  "metadata": {
    "subset_size": 5000,
    "max_items_in_memory": 5000,
    "total_items": 12000,  // Original number of items
    "significant_items": 8000,  // Number of genome-wide significant items
    "output_file": "/tmp/large_result_abc123.json"
  }
}

IMPORTANT:

• Always check the is_complete and output_file fields. If is_complete is false, only a subset of results is in items and the full result is saved to the file specified by output_file.
• For endpoints that process p-values (e.g., associations), total_items represents the original count, while total_items_aft_process represents the count after filtering.
• Study-related endpoints do not include p-value related metadata (significant_items).

Special Output Notes

• get_trait_associations may return a list of association IDs or, if the response format is unexpected, the raw association data structure.
• Some endpoints (notably those using the summary-statistics API) may return a single object in items if only one result is found.

Credits

This tool relies on the GWAS Catalog REST API and GWAS Summary Statistics API.

Please cite and credit the GWAS Catalog and each study when using this tool in your work.

License

This MCP server itself is licensed under the Apache License 2.0 - see the file for details.

This project uses the GWAS Catalog REST API and data provided by EMBL-EBI. Please ensure you cite the GWAS Catalog and the original studies when using this tool or its outputs. See the GWAS Catalog Terms of Use for details.

Acknowledgements

• GWAS Catalog
• FastMCP