Fig. 1

Flowchart of pipeline used to identify copy number variations (CNV). RO reciprocal overlap. *Only for the comparison of CNV between Pindel and DELLY. First, CNV were predicted separately by each tool within each sample. Second, only predictions supported by at least three reads and covered by less than 25% of their sequence of “N” gaps were retained. These filtered CNV were compared within the same sample to retain common CNV predicted by the different tools. CNV should share at least 90% of their sequence between two tools. Finally, CNV were identified by merging variants at 70% of reciprocal overlap across all samples