The presence of low quality samples during genotype calling may affect the cluster definitions and, therefore, the accuracy of genotype calls for high quality samples. This effect was demonstrated by Pluzhnikov et al. [2008] in a project that used the Affymetrix 5.0 array with Birdseed v2 calling by plate. Eight low quality samples were detected with high heterozygosity and “unusual patterns of relatedness”. These samples were all on one plate, which had a disproportionately high number of cases. False positive associations were found and these remained after simply removing the eight low quality samples. Re-calling of the remaining samples on the affected plate was necessary to remove the false positives.
