Comparison of THetA to CNAnorm and ABSOLUTE on simulated mixtures from real sequencing data. (A) Comparison of true and inferred tumor purity by THetA, CNAnorm and ABSOLUTE on simulated mixtures of DNA sequencing data from an acute myeloid leukemia sample and a matched normal sample. Gray dashed line indicates True Purity = Inferred Purity. Below each plot are the root mean squared errors (RMSEs) for each method. (B) Comparison of the number of copy number aberrations correctly predicted (defined as 50% reciprocal overlap in position and correct integral copy count) by each method for varying tumor purity and sequence coverage. Num TP is the number of true positive copy number aberrations predicted. In most cases, THetA outperforms both CNAnorm and ABSOLUTE. Similar results counting aberrations with correct position (with 50% reciprocal overlap) but allowing for difference between true and predicted copy number are in Additional file 1, Figure S7. RMSE: root mean squared error.
Oesper et al. Genome Biology 2013 14:R80 doi:10.1186/gb-2013-14-7-r80