PADME - Platform for Analytics and Distributed Machine Learning for Enterprises

PADME is our Distributed Analytics (DA) infrastructure that enables a paradigm-shift, in which we bring algorithms to the data instead of vice versa to enable privacy-preserving analysis on sensitive data. PADME leverages the Personal Health Train (PHT) concept, which is grounded on the FAIR principles from the GoFAIR initiative to enable Findability, Accessability, Interoperability and Reusability of healthcare data in a network of different institutions. In the PADME ecosystem, data owners are referred to as “station admins”, who stay in control and maintain their sensitive data in its original location, referred to as “stations”. Analytical tasks are referred to as “trains” and visit respective “stations” and only execute analytical computations after approval through the “station admins”. With PADME we propose a flexible solution for persistent privacy-related challenges and adherence to data protection requirements, especially in the healthcare domain, but also other domains requiring the analysis of distributed sensitive data. PADME’s architecture is containerized through docker and consists of a central service (CS), which manages train orchestration, operational logic, business logic and data management across the different containerized station services. The central service and station services are coupled through REST API connections and accessible via an intuitive user interface through a browser. PADME is continously further developed and evaluated in the scope of the Medical Informatics Initiative (MII) and the Horizon Europe project "BETTER".