Skip to main content
Figure 5 | Genome Biology

Figure 5

From: A general framework for analyzing tumor subclonality using SNP array and DNA sequencing data

Figure 5

In silico validation of CHAT performance. (A) Performance of sAGP inferences. Upper row: percent of error in estimated nb or nt, for the dominant (left) and subclonal sCNAs (right), as described in Materials and methods, Performance of sAGP inference. Middle row: the median absolute difference (MAD) between estimated and simulated sAGP values for sCNAs with correctly identified (nb, nt), or for all sCNAs (Bottom row). The psub =0 row of the lower-right and middle-right panels had zero error because when psub =0 there is only one clone in the tumor population and all subclonal sCNA segments have correctly estimated sAGP =0. (B) Performance of CCF inference. Shown are scatter plot of simulated and estimated CCF for four pdom - psub cases and two coverage values: Cov =50 (upper panels) and 100 (lower panels). (C) Comparison of CCF inference accuracy among different SNV categories: euploid vs. aneuploidy regions; and in the latter, between the dominant and the minor clones. Lastly, SNVs were divided by sCNA types. The tested case has the following parameter settings: pdom =0.9, psub =0.6, coverage =50, number of SNV sampled =4,000, number of sCNA sampled =200. ρ, Spearman’s correlation coefficient between the true and the estimated CCF values. MAD: median absolution difference between the true and the estimated CCF values.

Back to article page