As a more sensitive method may also be less specific, we also compared the performance of the methods with regard to the two kinds of false positives. We observed a very low false positive rate due to miscalled germline sites for all methods given sufficient depth (≥15x) in the matched normal (Fig. 4b). The false positive rates per megabase owing to miscalled reference sites (Fig. 4c) are comparable above 20x in both the STD configuration (median=10.2, range=0.7–20.1) and the HC configuration (median=1.0, range=0.2–3.1).
