Benefits of accurate imputations in GWAS

Abstract

Imputation methods have been suggested as an efficient way to increase both utility and coverage in genome-wide association studies, especially when combining data generated from different genotyping arrays. We aim to demonstrate that imputation results are extremely accurate and the association analysis from imputed data does not over-inflate the results. Instead imputation leads to an increase in the power of the dataset without introducing any systematic biases. The majority of common variants can be imputed with very high accuracy (r2>0.9) and we validated the accuracy of imputations by comparing actual genotypes from low-throughput genotyping assays against imputed genotypes. Imputation was performed using IMPUTE2 and the 1000 Genomes cosmopolitan reference panel, which results in about 38 million SNPs. After quality control and filtering we performed case-control associations with 3,159,556 markers. We show a comparison of results from genotyped and imputed data and also determine how accurate ancestry is determined by imputations.