We processed six samples in triplicate using 11 different array platforms at one or two laboratories. Each data set resulting from these experiments was analyzed by one or more CNV calling algorithms. The DNA samples originate from HapMap lymphoblast cell lines and were selected based on their inclusion in other large-scale projects and their lack of previously detected cell line artifacts or large chromosomal aberrations. An overview of the platforms, laboratories and algorithms is shown in Table 1, with additional details of the arrays and their coverage in Supplementary Tables 1 and 2 and Supplementary Figure 1. We assessed the experimental results at three different levels. First, we obtained measures of array signal variability based on raw data before CNV calling. Then, the data sets were analyzed with one or more CNV calling algorithms to determine the number of calls, between-replicate reproducibility and size distribution. In the third step, we compared the CNV calls to well-characterized and validated sets of variants, in order to examine the propensity for false-positive and false-negative calls and to estimate the accuracy of CNV boundaries.
