A common step in genome-wide association studies is to impute missing genotypes not directly ascertained by the genotyping platform used. The accuracy of this imputation depends on the reference panel(s) selected, and how the results are combined. Frequently, the most suitable panels are hosted by specific sites and cannot be relocated (e.g. TOPMed, H3Africa).

The Federated Genotype Imputation GIF Project, led by eLwazi, is exploring this challenge. eLwazi is exploring the following use case: a user has a set of 1,000 samples genotyped on a SNP microarray. They want to impute the unobserved genotypes using multiple reference panels hosted at different locations. After submitting to the frontend hosted at eLwazi, these study genotypes are submitted to the chosen services for imputation. Once these tasks are complete, the results are collated for the user. It is important to note that this solution does not solve the problem of restrictions on copying target datasets. The user is still submitting their genotypes to these services hosted elsewhere. The frontend service will make this clear to the user.

The outcome will be a frontend service that accepts target genotypes, runs imputation at the multiple selected remote services, then collates and presents the results. The value will be in a single frontend to imputation instead of the current pattern of taking data to the various frontends. By adopting GA4GH standards where possible, components of this project will be reusable in other eLwazi and non-eLwazi projects (most notably the imputation services).

Interested in learning more? Reach out to the Technical Leads Ayton Meintjes and Mamana Mbiyavanga.