Figure 6:

Performance of other datasets (A: SureSelect, B: TruSeq, C: Nextera) when analyzed using parameters optimal for Nimblegen data (opt) compared to default settings (default). Nimblegen-optimized parameters are frequently unsuitable on other datasets. XHMM is severely compromised by the Nimblegen settings on all datasets. SureSelect and TruSeq data produce no XHMM CNV calls, while both sensitivity and precision are poorer in Nextera data. CoNIFER and cn.MOPs both gain in sensitivity but by a much smaller proportion and with a larger inflation of false-positive calls than with Nimblegen data.