Given the well-understood requirements for large sample size for GWAS data, we used simulations to investigate the power of our approach. We generated pairs of eQTL/biomarker datasets assuming a shared causal variant. We varied two parameters: the sample size of the biomarker dataset and the proportion of the biomarker variance explained by the shared genetic variant. We set the proportion of the eQTL variance explained by the shared variant to 10% and we used the original sample size of the liver eQTL dataset described herein [16]. Text S1 contains a description of the simulation procedure.
