Federated learning

NextGen

Genome-centric multimodal data integration for personalised cardiovascular medicine

• Federated learning
• Healthcare
• Ongoing

Aim

Personalised prevention, diagnosis, monitoring and treatment of cardiovascular disease require artificial intelligence models that use multimodal patient data, and genomic data in particular. Privacy and governance requirements, multiple standards, heterogeneous data formats and the volume of the data make this integration difficult. NextGen develops tools that remove these barriers for several cardiovascular use cases, including secure federated analytics extended to genomic computation, more effective federated learning over distributed infrastructures, scalable curation of genomic data and variant prioritisation. The tools are demonstrated in real-world pilots in a network of five clinical sites. The project contributes to the European 1+ Million Genomes initiative and to the European Health Data Space.

Our contribution

IDSIA contributes to two objectives of the project:

• tools for the prediction, prevention, diagnosis, monitoring and treatment of cardiovascular disease from multi-omic and multimodal data;
• better use of genomic data through data integration and workflow tools, including federated learning algorithms that overcome the barriers to the integration of health data.

Publications

• Technical and legal aspects of federated learning in bioinformatics: applications, challenges and opportunities

  D. Malpetti, M. Scutari, F. Gualdi, J. van Setten, S. van der Laan, S. Haitjema, A.M. Lee, I. Hering, F. Mangili

  Frontiers in Digital Health, 7, 1644291, 2025Code

Software

• Federated learning in bioinformatics, examples

  Example projects with the Flower framework (weakly expressed genes, minor allele frequencies, secure summation) that accompany a review of the technical and legal aspects of federated learning in bioinformatics.

  GitHubPaper