Case and control CNVs were pooled and clustered into CNV events (CNVEs) as previously described 20. 587 CNVEs with a carrier frequency >1% were defined as common CNVEs. Each CNVE was analyzed using the R package CNV tools which implements a likelihood ratio test that models the distribution of per sample CNV measurements as a Gaussian mixture and compares the goodness of fit with or without association to affected status. Test p values could be obtained for 481 of the 587 common CNVEs, whereas the rest failed to run through CNV tools due to poor genotype clustering. For further analysis of NEGR1 deletions, association tests were performed using Fisher’s exact test based on discrete genotype calls, as the distribution of CNV measures of both cases and controls were in distinct and non-overlapping copy number clusters.
