Being able to share large amounts of sensitive clinical and genomic data across several institutions is crucial for precision medicine to scale up. Unfor- tunately, existing solutions only partially address this challenge and are still unable to provide the strong privacy and security guarantees required by regulations (e.g., HIPAA, GDPR). As a result, currently only very limited datasets of non-sensitive and moderately useful information can be shared. In this paper, we introduce MedCo, the first operational system that enables an investigator to explore sensi- tive medical information distributed at several sites and protected with collective homomorphic encryption. MedCo builds on top of established and widespread technology from the biomedical informatics community, such as i2b2 and SHRINE, and relies on state-of-the-art secure protocols for processing encrypted distributed data and complying with regulations. As such, MedCo can be easily adopted by clinical sites thus paving the way to new unexplored data-sharing use cases. We tested MedCo in a real network of three institutions (EPFL, UNIL and CHUV) by focusing on an oncology use-case with real somatic mutations and clinical tumor data. The relatively low overhead introduced by MedCo shows that it represents a concrete and scalable solution for sharing privacy-conscious medical data.
PDF