Saliva was collected from participating subjects and DNA was extracted, purified and genotyped on an Illumina human 1 M-duo SNP assay resulting in 1,199,187 genotyped loci. CNVs were detecting using the log-R intensity ratio (LRR) and the beta allele frequency of each autosome locus in a process described in (Chen, Jingyu, Boutte, & Calhoun, 2011). Briefly, after an initial outlier and principal component correction to remove confounding experimental factors, the intensity and allele frequency data was segmented using both a circular binary segmentation algorithm (Olshen, Venkatraman, Lucito, & Wigler, 2004) and hidden Markov model based algorithm (Wang et al., 2007). Only those regions where both algorithms agreed were included in the final calls. Each copy number call was coded either 0,1,2,3,4 for homozygous and hemizygous variations with 2 representing the neutral case. Regions showing more than 1% incidence frequency of CNV calls were selected for further analysis resulting in a total of 253 CNV regions.
