Fig. 3

Results on high-grade serous ovarian cancer OV2295 [27]. a Allele-specific copy number profiles for CNRein, SIGNALS, CHISEL, and Alleloscope on \(n=617\) cells. Unlike other methods, Alleloscope did not predict WGD for any cell (as indicated by ploidies above 3.0). b Allele-specific copy number profiles of chromosome 1, showing that while CNRein retains the ability to detect small CNAs also detected by SIGNALS. c SIGNALS and CNRein have a substantially closer fit to the raw read depth data than alternative methods. d The size of the 10 largest clones for each method, showing larger clones for CNRein and CHISEL than SIGNALS and Alleloscope. e CNRein’s predictions resulted in a more parsimonious phylogeny than SIGNALS or CHISEL, while Alleloscope resulted in the lowest parsimony score. f, g VAF values on the copy number 1, 2 for CNRein and Alleloscope. For CNRein but not Alleloscope, VAFs are concentrated around 1/3 and 2/3 (indicated with black vertical lines) as expected. h Log likelihood ratios (LLR) of truncal SNV support on bootstrapped replicates between our method and the alternative methods SIGNALS, CHISEL, and Alleloscope. CNRein outperforms CHISEL (LLR 3,418.66 and \(p < 10^{-5}\)) and especially Alleloscope (LLR 12,635.36 and \(p < 10^{-5}\)) while being slightly outperformed by SIGNALS (LLR \(-1{,}033.42\) and \(p = 0.0051\))