In order to provide a more direct comparison of imputation accuracy between the MTFS and 1000 Genomes reference sequences, we compared imputation accuracy using dosage R2, the squared Pearson correlation between the minor allele genotype count from rare variants on the exome array and the imputed dosage. In this comparison, we restricted phasing and imputation to Chromosome 20 from the 660W-Quad only. We purposefully excluded all variants from the exome array during phasing and imputation, imputed all the exome chip variants that were discovered through sequencing, and then evaluated the accuracy of that imputation. We restricted this subsample to individuals with European ancestry, just as we have throughout this special issue in all association analyses. Finally, we included only one member of each MZ twin pair, because the other member is entirely redundant for purely genetic analyses such as imputation (recall the genotypes are copied from the genotyped twin to his/her co-twin). This resulted in a subsample of 6,610 individuals of European ancestry for 1,369 SNPs, for the sole purpose of evaluating imputation accuracy for rare variants.
