Cohort studies investigating aging and dementia require APOE genotyping. We compared directly measured APOE genotypes to 'hard-call' genotypes derived from imputing genome-wide genotyping data from a range of platforms using several imputation panels. Older GWAS arrays imputed to 1000 Genomes Project (1KGP) phases and the Haplotype Reference Consortium (HRC) reference panels were able to achieve concordance rates of over 98% with stringent quality control (hard-call-threshold 0.8). However, this resulted in high levels of missingness (>12% with 1KGP and 5% with HRC). With recent GWAS arrays, concordance of 99% could be obtained with relatively lenient QC, resulting in no missingness.
Keywords: Alzheimer’s disease; ApoE receptor; cohort studies; computational biology; genetic association studies.