GA4GH Connect - Key takeaways and next action items - GA4GH in Real World
Reggan Thomas
Hi all,
Thank you for joining us for the "GA4GH in Real world " session at the GA4GH October Connect meeting 2025!
We've updated the agenda document with the following key takeaways and next action items from the session.
Key takeaways:
1. Real-World Data Sharing: IPCHiP Case Study
- The International Precision Child Health Partnership (IPCHiP) exemplifies borderless data sharing using GA4GH principles.
- Key challenges include managing on-prem vs. cloud environments, ensuring data security, and providing metadata access.
- Emphasis on transparency in both successes and setbacks when adopting GA4GH standards.
2. Federated Variant-Level Matching and Imputation Platform
- A federated imputation platform, presented by Mamana, connects multiple imputation services through one interface, improving accuracy and efficiency.
- Built using GA4GH standards, currently supporting Michigan imputation software and multiple reference panels.
- Highlights the need for better integration between DRS and WES.
- Discussions emphasized regional engagement (e.g., in Uganda and Kenya) and starting with small-scale implementations.
3. Pacific Analytics Toolkit and Cloud Components
- Pacific Analytics demonstrated an open-source, framework-agnostic toolkit for building interoperable platforms using GA4GH standards.
- The system supports data discovery, workflow execution, and repository integration.
- Encourages collaboration across groups and integration of GA4GH standards with existing infrastructure.
4. Minimal Starter Kit for Interoperability
- Proposal for a “minimal starter kit” to deploy real-world interoperable systems beyond demo environments.
- Focuses on essential components: authentication, authorization, data models, APIs, storage, logging, and auditing.
- The approach involves defining scientific use cases, aligning on relevant standards, and deploying public data demos with stakeholder input.
5. Federated Machine Learning (FML) and Community Collaboration
- Federated ML discussed as a privacy-preserving alternative to centralized models.
- Challenges include data preparation, access permissions, and authentication/authorization.
- A new Cloud, ML, and AI Working Group formed to develop frameworks supporting federated ML across GA4GH standards.
- Emphasis on community engagement and addressing legal, ethical, and security aspects in federated environments.
Next steps:
1. Enhance Standards Integration
- Improve interoperability between DRS, WES, and other GA4GH APIs to support federated analysis platforms.
2. Advance the Minimal Starter Kit
- Define scope, core components, and priority standards.
- Develop and test a prototype deployment using public data.
3. Expand Federated Compute and ML Initiatives
- Continue work within the Cloud, ML, and AI group to define use cases and build a shared framework.
- Ensure strong links with security, ethics, and governance workstreams.
4. Promote Community Adoption
- Organize regional workshops (e.g., in Africa) to raise awareness and support small-scale pilot implementations.
- Foster collaboration among groups like Pacific Analytics and federated data initiatives.
5. Document and Share Learnings
- Publish case studies (e.g., IPCHiP, Federated Imputation) demonstrating real-world adoption of GA4GH standards.
- Encourage sharing of implementation experiences — both challenges and successes — to build community confidence.
If you missed out on the Connect sessions, don't worry! You can catch up by watching the meeting recording here (passcode: .P#Rgz0E)and reviewing the notes here. And if you have any further questions or want to continue the discussion, feel free to hop onto the slack channel #cloud_ws
Please take a moment to complete our Feedback Survey: https://forms.ga4gh.org/t/qjHm9R1EGmus
Feel free to reach out if you have any questions. Looking forward to staying connected!
On behalf of
Brian and Venkat
Please note there is no obligation to respond to emails from me outside of your typical working hours.