Dependency of accuracy on coverage represented by the number of reads. N reads (N = 500, 1000, 2000, 4000, 8000, 16,000) were randomly selected 10 times from the original data and both methods 2SNV and PredictHaplo were applied. For N = 500 and N = 1000, PredictHaplo gave results only in 5 and 9 runs, respectively. Each dot represents the reconstructed frequency of the clone in the respective runs.