GDI D8.9 - Distributed analysis and federated learning PoC
-
1.
Barcelona Supercomputing Center
-
2.
Central European Institute of Technology
Contributors
-
1.
German Cancer Research Center
-
2.
National Institute of Health Dr. Ricardo Jorge
-
3.
Instituto Superior Técnico
-
4.
University of Maribor
-
5.
Masaryk University
-
6.
Barcelona Supercomputing Center
-
7.
University of Tartu
-
8.
CSC, IT Center for Science Ltd.
- 9. ELIXIR Hub
- 10. Danish National Genome Center
Description
This deliverable describes the design, implementation, and testing of a demonstrator for federated analysis within the European Genomic Data Infrastructure (GDI) project. It showcases a practical execution of a privacy-preserving Genome-Wide Association Study using a distributed architecture. Each participating national node (CZ, DE, PT, ES, SI, FI, EE) ran analyses locally, preserving data sovereignty and complying with legal constraints such as the GDPR. This demonstrator was built upon prior strategic and technical evaluations (previous deliverables and discussion from both WP7 and WP8) of federated technologies, validating the technical feasibility of GA4GH-compliant TES workflows. It leverages tools such as Snakemake, Docker, and MinIO, with Funnel serving as the TES endpoint. This prototype, using synthetic1 data previously distributed across the national nodes and generated for this demonstrator, paves the way for federated learning and cross-border analytics at scale.
Files
202504 - GDI_D8.9 Distributed analysis and federated learning PoC.pdf
Files
(3.4 MB)
| Name | Size | Download all |
|---|---|---|
|
md5:2cb35ed195b37105b7d4f0f437c5c260
|
3.4 MB | Preview Download |
Additional details
Funding
- European Commission
- European Genomic Data Infrastructure (GDI) 101081813