Federated Learning for multi-omics: a performance evaluation in Parkinson's disease

Area of Science:

• Computational biology
• Genomics
• Machine learning

Background:

• Machine learning (ML) application in omics is limited by data access.
• Federated learning (FL) offers a solution for collaborative, privacy-preserving data curation.

Approach:

• Compared FL models with centrally trained ML models for multi-omics Parkinson's Disease prediction.
• Evaluated performance using AUC-PR, analyzing the impact of sample dispersion.

Key Points:

• FL model performance closely mirrors centrally trained ML models.
• The most effective FL model achieved an AUC-PR of 0.876 ± 0.009.
• Sample dispersion within a federation significantly influences model performance.

Conclusions:

• FL is a viable approach for multi-omics studies, overcoming data access limitations.
• Open-source FL frameworks can be effectively implemented for collaborative research.
• Further research is needed to optimize FL strategies for omics data challenges.