Figure 2

Measuring sensitivity, specificity, and Matthews correlation coefficient (MCC) for optimal choice of resolution threshold and base quality threshold. Analysis was done averaging sensitivity, specificity, and MCC across three 0.1% synthetic mixture DNA replicates sequenced on a GAIIx platform. (A) With a base quality threshold of 30, the specificity increases slightly from 0.999 to 1 as the resolution threshold is increased from 0.01% to 0.1%. More drastically, the sensitivity decreases at a threshold greater than 0.05% from near 1 to 0.857 at 0.1%. The MCC statistic that combines these two measures and adjusts for unequal distribution of true positives and true negatives peaks a range of 0.02% - 0.07%, indicating an optimal resolution threshold. (B) As the base quality threshold is increased from 0 to 30, RVD sensitivity becomes less dependent on resolution threshold. Thus, a higher base quality threshold may improve resolution of the algorithm.