Fig. 2

CNAtra solution for multi-level CNA detection in low-coverage data. a Schematic representation of multimodal distribution under different ploidy assumptions. Dotted red lines denote the boundaries of CN states (CN intervals). b A hierarchical computational framework for detecting both LCVs and FAs. The CNAtra approach includes smoothing of RD signal using Savitzky-Golay filter (black line) (top panel), followed by detection of IBs with distinct CN states (middle panel) and then identification of FAs inside each IB (bottom panel). c Effects of tuning the amplification/deletion thresholds (Th) on the detection of false positive FAs in a region in Chr 1 (195–203 Mb) using subsampled data of A427 cell line. The red and black triangles represent the true positives and false positives respectively. d Calibration of CNAtra parameters such as resolution (top panel), amplification threshold (middle panel) and deletion threshold (bottom panel) using high-coverage WGS datasets. The solid black line represents the fitted regression line using negative exponential model