The second problematic situation for imputation is where cases and controls are genotyped on different platforms. This is problematic because imputation error can vary between cases and controls, causing increased rates of false positives in association studies. There is no known method for effectively filtering the poorly imputed SNPs from the well imputed SNPs on different platforms. Although this situation has been avoided by researchers, it is an important application. Large studies such as Wellcome Trust and the NIMH GAIN samples use common controls that could be used in other studies to gain power [1], [14]. But, if the primary datasets were genotyped on a different platform, imputation is necessary.
