Figure 2. Illustration of the procedure to estimate imputation accuracy.

We used a drop one-out crossvalidation approach. For the imputation step each subject was removed from the reference panel in turn, and this subject’s exome sequence SNPs were then imputed using either the 1000 Genomes reference panel alone or in conjunction with a second local reference panel. All subjects’ imputed allelic dosages were then compared with the exome sequence genotype data (“gold standard”).