Research Agents Demonstration Using GA4GH APIs

Background

Over the last several years we have seen the rise of development and use of Artificial Intelligence (AI) and specially Large Language Models (LLMs) to accelerate discovery and use of data and software. A central insight is that the same GA4GH standards that enable human researchers to discover data, access it, and run workflows against it, are also the natural building blocks for autonomous AI agents acting on behalf of researchers. 

The Research Agents GIF project aims to demonstrate end-to-end scenarios in which an AI agent — acting on behalf of a researcher — uses the GA4GH implementation registry and GA4GH-compliant APIs to discover available data and compute resources, construct an analysis plan, and execute it. The goal is a working, reproducible reference implementation that the broader community can learn from and extend.  

Ongoing Work

The team is developing reference AI agent implementations that demonstrate how researchers can discover, access, and analyse data using GA4GH standards and APIs. Current areas of work include:

• Query the registry to discover what services are available across participating sites
• Reason about which services and datasets are relevant to a researcher’s stated question
• Obtain authorisation to access controlled data
• Resolve DRS URIs to retrieve or reference the relevant datasets
• Identify and retrieve an appropriate workflow from a TRS-compliant registry (e.g. Dockstore)
• Submit the workflow for execution via WES or TES at a participating site
• Monitor execution and return results to the researcher

How to participate

Interested in helping build AI-ready genomic infrastructure? Learn more, contribute use cases, participate as a deployment site, or help develop agent tooling using GA4GH standards.

Join the Slack Channel. 

Visit our GitHub repository to view open issues, provide feedback, or propose new ideas: https://github.com/ga4gh/gif-project-ai-agents 

References

Products used in this GIF project 

• Service Registry
• Beacon v2
• Data Connect
• Passports and Visas
• Data Repository Service (DRS)
• Tool Registry Service (TRS)
• Workflow Execution Service (WES)
• Task Execution Service (TES)
• Refget