Preprint publication

FedRBE – A decentralized privacy-preserving federated batch effect correction tool for omics data based on limma

Yuliya Burankova, Julian Klemm, Jens J. G. Lohmann, Ahmad Taheri, Niklas Probul, Jan Baumbach & Olga Zolotareva · arXiv · December 2024

Batch effects represent a major challenge for distributed biomedical and omics data analysis, especially in privacy-sensitive healthcare settings where patient data cannot easily be centralized.

In this work, the authors present FedRBE, a decentralized federated implementation of limma’s removeBatchEffect method developed for the FeatureCloud platform. The tool enables collaborative correction of omics datasets across institutions without requiring direct data sharing while supporting missing values and automated workflows.

The study demonstrates that FedRBE achieves results comparable to centralized approaches with negligible differences, highlighting its potential for secure multi-center biomedical research and privacy-preserving AI applications.

Read full preprint

Why this matters for Microb-AI-ome

Privacy-preserving federated learning is a central technological pillar of Microb-AI-ome. This publication contributes directly to the development of secure AI infrastructures for distributed biomedical and microbiome research, enabling collaborative analyses across institutions without compromising sensitive patient data. The work also highlights the growing role of federated omics analysis in future precision medicine applications.