Jan Baumbach1,2
1Institute for Computational Systems Biology, University of Hamburg, Notkestrasse 9, 22607 Hamburg, Germany
2Computational BioMedicine lab, Institute of Mathematics and Computer Science, University of Southern Denmark, Campusvej 1, 5000 Odense M, Denmark
jan.baumbach [at] uni-hamburg.de
Abstract
Artificial intelligence (AI) offers game-changing opportunities to healthcare. However, it also harbors risks to patient privacy in particular when dealing with sensitive clinical data stored in critical healthcare IT infrastructure. Specifically, data exchange over the internet is perceived insurmountable, posing a roadblock hampering big-data-based medical innovations.
We created a novel AI platform, the FeatureCloud AI app store that is based on the idea of federated learning where only model parameters are communicated. To maximize privacy, sensitive datasets remain stored locally and are analysed behind safe firewalls to assure the high standards in data privacy in order to (by design) comply with the strict GDPR.
We will exemplarly investigate the power of FeatureCloud apps for decentralized (1) genome-wide association studies (GWAS), (2) gene expression data mining, and (3) time-to-event data analytics to demonstrate how FeatureCloud may enhance worldwide collaboration, accelerate innovation, and democratize scientific data usage. We show that apps developed in FeatureCloud can produce highly similar results compared to centralized approaches and scale well for an increasing number of participating sites.
FeatureCloud is a no-code platform for federated learning apps having the potential to vastly increase the accessibility of privacy-preserving and distributed data analysis in biomedicine and beyond.
Keywords: bioinformatics, data mining, federated learning
Acknowledgement: This project has received funding from the European Union’s Horizon2020 research and innovation programme under grant agreement No 826078.
