Fig. 5

The performance of CNETML on relative copy number data. The relative copy numbers were obtained by using the rounded mean copy number as baseline and direct rounding. There are 500 simulated datasets in total, which are divided into four groups by the types of WGDs. The number of datasets in each group is shown in brackets. MEDICC2 was excluded when comparing branch score distance because the branch length in a tree built by it has a different meaning (the number of events between CNPs of two nodes based on CNT model) and it is hard to compare fairly. Ninety-one outlier data points with values larger than the maximum of x-axis on datasets with subclonal WGDs are excluded in the plot of mutation rates for better visualization of the majority data. Box plots have the same interpretations as those in Fig. 3