Secure large-scale genome-wide association studies using homomorphic encryption

We propose a toolbox of statistical techniques that leverage homomorphic encryption (HE) to perform large-scale GWASs on encrypted genetic/phenotype data noninteractively and without requiring decryption. We reformulated the GWAS tests to fully benefit from encrypted data packing and parallel computation, integrated highly efficient statistical computations, and developed over a dozen cryptoengineering optimizations. Our HE solution is >30× faster than the cutting-edge multiparty computation method, in contrast to the claim that HE is not viable for large-scale GWASs. Our approach is thus a significant advance both in methodology and application, empowering large-scale cross-institution collaboration, patient-driven research, and crowdsourced genomics.