Privacy-preserving Polygenic Risk Scoring using Homomorphic Encryption

Abstract

The availability of direct-to-consumer genetic testing services and genome sequencing data bring novel opportunities for applications like genomic risk scoring where a polygenic disease risk score is calculated considering the statistical distribution of the disease associated SNPs. Nowadays, various websites are offering polygenic risk score estimations for various complex diseases. However, these services require the upload of the genomic data to their sites, which is a fairly sensitive personal data. Since, genome data uniquely identifies a person, anonymization is not sufficient alone and may become a threat in the long run. A potential solution is the use of cryptographic techniques along this goal. We propose to deploy homomorphic encryption, a technique which enables to do computation in encrypted data, for a web server providing polygenic risk score estimation. We implemented a proof-of-concept software to measure the performance of such a service with current technology. We also developed a GUI which facilitates the usage of homomorphic encryption for non-technical users. We conclude that recently developed homomorphic encryption libraries enable practical privacy-preserving genomic risk scoring services. Homomorphic encryption is becoming a strong alternative for practical secure privacy-preserving personalized medicine applications.