Detection of CNV from exome or targeted-capture sequence data presents unique challenges due to the increased GC bias inherent to targeted-capture data, and the discontinuous nature of the coverage profile. For custom targeted-capture sequence data, in which large regions (>1 kb) are targeted, CNV can be detected within single samples (no controls) following correction for GC-content and edge effects (41). However, because of the small size of targets in typical exome-capture data, many current algorithms for CNV detection require either a paired normal sample or a panel of population controls. CONTRA, one such method for CNV detection from exome data, first calculates the tumor/ normal coverage ratio exome-wideband then employs a normal approximation to detect CNV at the exon level (42). Finally, exon-level deletions or duplications are merged into larger CNV using circular binary segmentation (CBS). CoNVEX detects CNV using a similar strategy, first denoising the coverage ratio using a discrete wavelet transform, and then identifying copy gains and losses via a hidden Markov model (43). A third, similar, approach is taken by ExomeCNV, which segments the exome into regions
