Resources

Preconstructed Indexes

Access MetaGraph indexes via AWS S3 for local analysis using the MetaGraph command line tool. All indexes are available for download from AWS Open Data. Install the AWS CLI following the installation guide (supports Windows, macOS, Linux).

Example commands:

# List available objects in a bucket
aws s3 --no-sign-request ls s3://metagraph/refseq/

# Download a specific file
aws s3 --no-sign-request cp s3://metagraph/refseq/file.dbg .

# Sync an entire directory
aws s3 --no-sign-request sync s3://metagraph/refseq/ ./local-refseq/

The --no-sign-request flag indicates public access without AWS credentials.

Once downloaded, the indexes can be queried locally with the command line interface and Python API (see Query index).

For more information about the indexes and deployment manuals, see https://github.com/ratschlab/metagraph-open-data.

MetaGraph Online

MetaGraph Online provides a web interface for searching DNA, RNA, and protein sequences across various public archives indexed with annotated de Bruijn graphs.

Web Interface Features:

  • Search up to 10 sequences per query (maximum length 50k)

  • Support for both exact matching and sensitive alignment

  • Access to various indexes (for current coverage see Databases)

    • Microbe, Fungi, Plants, Metazoa (including Human/Mouse)

    • Reference/assembled sets (RefSeq, UHGG, Tara Oceans)

    • Protein databases (UniParc)

  • Export results as CSV or JSON

  • Shareable links with 48-hour expiration (6 months with locked searches)

API Access: The production API endpoint is available at: https://metagraph.ethz.ch:8081/search

For detailed API documentation and examples, see the MetaGraph Online Help page.

Note

Python API described in the API documentation is used in the internal implementation of the service and can also be used to query indexes hosted locally. To query indexes hosted publicly, refer to MetaGraph Online Help.

Web Interface Limits: Maximum 10 sequences per query, 50k bases, results expire after 48 hours (6 months with locked searches).

Performance Tips:

  • Use exact matching for high identity matches (fast)

  • Use alignment mode for divergent/noisy sequences (slower)

  • For larger jobs, use the API (rate limits apply for queued jobs) or download Preconstructed Indexes and query locally

  • Split large inputs to avoid timeouts

Contact & Support

For issues or feature requests: