Performance evaluation of the HSLM algorithm for detecting CNVs in synthetic chromosomes. (a), (b) ROC curves comparing the sensitivity and specificity of the HSLM and CBS algorithms in the detection of one-copy (a) and three-copy CNVs (b). (c), (d) Comparisom of the HSLM and CBS algorithms when analyzing synthetic chromosomes with different numbers (g = [ 1,2,3,4,5]) of one-copy (c) and three-copy (d) genes. (e), (f) Performance of the HSLM (e) and CBS (f) algorithms in detecting the correct breakpoint position. The x axis is the distance between the predicted and the correct position. The y axis is the percentage of breakpoints predicted at a given distance from the correct position. (g), (h), (i), (j) TPR and FP plots for different values of the DNorm parameter versus exon number in the segmented region. (g) and (h) show TPR and FP when analyzing one-copy regions. (i) and (j) are TPR and FP when analyzing three-copy regions. Each curve point was obtained by averaging across 5,000 simulations (1,000 synthetic chromosomes for g = [ 1,2,3,4,5]).CBS, circular binary segmentation; CNV, copy number variant; FPR, false positive rate; HSLM, heterogeneous shifting level model; TPR, true positive rate.