Following substantial improvements in phasing software and haplotype reference panels for populations of Caucasian ancestry, a second generation of imputation was performed for the autosomes in March, 2017 on the Michigan Imputation Server (MIS), using Minimac3, the Haplotype Reference Consortium (HRC) reference panel (v.1.1), and Eagle (v2.3) phasing software. Pre-imputation quality control identified 23 subjects from the Affy6.0 platform dataset (initial n=1709) and 3 subjects from the Illumina platform dataset (initial n=384) with high proportions of missing genotypes (>0.5) for at least one 20MB region of the genome, yielding final sample sizes of n=1686 and n=381 imputed using MIS. After imputation, these datasets were merged into a single n=2067 fileset. The number of variants imputed with high confidence (INFO score>0.8) was >11.2 million, representing a large increase over the 1000 Genomes Phase 1 imputation dataset in the number of high quality variants available for analyses. Comparisons of subject-level genotype discordance for overlapping SNPs between the 1000 Genomes Phase 1 imputation, the MIS imputation, and whole genome sequencing (WGS) found an average discordance of 0.7% for MIS and 2.5% for 1000
