We show that targeted tumor sequencing is sufficient to infer common germline variation, with applications to GWAS, polygenic risk scores, and genetic ancestry. We benchmark multiple methods and investigate potential biases due to somatic alterations. We advocate that this is as a way to greatly increase the sample size of studies with germline and somatic data in the same individuals. We have already applied this method to study germline predictors of immunotherapy response (Luo et al. Clin Cancer Res) and ancestry-somatic associations (Carrot-Zhang et al. Cancer Discov).
A full pipeline and deployable workflow are available in a repository.
Figure: Schematic of the off-target imputation process