Detection of inappropriate samples in association studies by an IBS-based method considering linkage disequilibrium between genetic markers

An association study is a popular study design to identify susceptibility genes for common complex diseases. In such a study, the presence of inappropriate samples, such as those derived from close relatives or showing DNA contamination, causes an inflation of type I error or a decrease of power. Here we propose an identity-by-state (IBS)-based detection method of inappropriate samples taking linkage disequilibrium (LD) into consideration. The test statistics is the mean of the proportion of alleles that are shared identical by state at each single nucleotide polymorphism (SNP) between each sample pair in an association study. A covariance of the number of shared alleles between two SNPs is introduced to consider LD. We show that type I error and power are estimated accurately in computer-simulated data, and that if the number of SNPs analyzed is small, the performance of detection of inappropriate samples is superior to the previous method in simulated LD. An application to real association study data showed that accuracy in estimating the distribution of test statistics improved if LD was considered. Sample pairs considered to be siblings were detected. These results suggested that an LD-considered IBS-based detection method is useful in identifying inappropriate samples in an association study.