In this study we introduced a new method for genotype imputation in large association studies. Our method, IMPUTE version 2, follows a flexible inference framework that uses more of the information in the data than many comparable methods, thereby improving accuracy, while remaining computationally tractable on large datasets. This approach is well-suited to the kinds of datasets that will become available in next-generation association studies: it can handle large reference panels, including ones with unphased and incomplete genotypes, and it can also integrate multiple reference panels containing different sets of SNPs.
