One normal control (No. 1) was run and analysed three times on both the Human-1 and HumanHap300 platforms: in both cases the genotyping data were highly concordant (>99%). Despite this, the BeadStudio LOH+ analysis suggested some discordant events among the three replicates (Supplementary Table S1). With QuantiSNP, we could combine the Human-1 and HumanHap300 datasets (which is not currently possible in BeadStudio). Using a log Bayes Factor of 30 and a characteristic length of , we consistently identified two CNV events (one very small homozygous deletion on chromosome 1 and one duplication on chromosome 12) in all the three replicates of sample 1 using the combined dataset (Supplementary Table S2A). From our calibration study, we found that these settings corresponded to a false positive rate of less than one false CNV event call per 100,000 SNPs. This setting was chosen to be deliberately stringent in order to limit the number of CNV event calls made since we are unable to independently validate the existence or otherwise of large numbers of putative CNVs.
