For duplicate sample pairs, the median discordance rates (discordant calls per SNP) are 7 × 10−5 for Lung Cancer (33 pairs, Illumina HumanHap550 array) and 2 × 10−4 for Addiction (60 pairs, Illumina Human1M array), so the genotyping error rates are on the order of 10−4. The corresponding mean completion (call) rates are very high: 99.8% for Lung Cancer and 99.7% for Addiction samples. For the T2D projects run on the Affymetrix 6.0 array, study samples were not duplicated, but multiple replicates of a single HapMap control sample (NA12144) provide discordance rate estimates of 4 × 10−3 for NHS and 1 × 10−3 for HPFS. The corresponding mean completion rates are 99.6% for NHS and 99.7% for HPFS. It appears that the error rate is about an order of magnitude larger for the Affymetrix 6.0 than for the Illumina 1M arrays, although different sample sets and other factors could affect these results.
