Fig. 2

Evaluate the performance of negative-defining methods in a simulated dataset. From bottom to top: we simulated four different scenarios containing varying tumor cell fractions from 90, 20, 5 to 1%. Each scenario was independently simulated three times (referred to as measurements) to mimic multiple sampling. Only mutations that are positive in at least one of the three measurements after simulation were included. X-axis: different negative-defining methods including MSN using two thresholds (p < 0.01 and p < 0.05) and UMC using four thresholds (minimum coverage for non-positive samples to be considered as negative: 20X, 50X, 200X and 300X). Y-axis: percent of defined mutation statuses by type (Unknown: non-positive but the coverage was too low to be considered as negative; FN false negative, TN true negative, FP false positive, TP true positive). Please note that the current negative-defining methods do not affect positive mutation statuses (TP and FP)