Imputation server: next generation genotype imputation service. C. Fuchsberger1, L. Forer2, S. Schönherr2, F. Kronenberg2, G. Abecasis1 1) Ctr Statistical Genetics, Univ Michigan, Ann Arbor, MI; 2) Division of Genetic Epidemiology, Department of Medical Genetics, Molecular and Clinical Pharmacology, Innsbruck Medical University, Innsbruck, Austria.

   Genotype imputation is a key step in the analysis of genome-wide association studies (GWAS). However, imputation into large GWA studies requires expertise and substantial computational resources. Moreover, although upcoming mega reference panels, such as the 30k panel from the Haplotype Consortium, will improve imputation of rare and less common variants, but cannot always be shared broadly, due to consent and privacy restrictions on the original samples. To keep imputation broadly accessible, we designed a web service (called Imputation server) that imputes GWAS data following the MapReduce paradigm without directly sharing the reference panel. To ensure a high level of data sensitivity we have implemented several strategies. All interactions with the server are encrypted. After the imputation process is completed, results are encrypted with a one-time password and are only kept for a few days on our server. Input data is deleted, as soon it is no longer needed. We protect non-public available reference panels by preventing direct access and pseudo imputations geared to unveil the identity of reference panel members. To make this service highly scalable, we have re-engineered the core algorithms in our imputation engine, resulting in a speed-up of ~20x compared to our previous implementation. Our imputation server accepts phased and unphased GWAS genotypes and performs several quality checks, such as strand orientation, variant coding, file integrity, minor allele frequency, and sample and variant missingness. Our service can handle imputation of ~1,500 unphased genomes per day and can easily be scaled up. The service can be accessed for free at https://imputationserver.sph.umich.edu.

You may contact the first author (during and after the meeting) at