Fig. 2

Verification Selection Experimental Design. Verification candidates were selected from somatic mutation calling results of multiple algorithms run on three in silico tumours (IS1, IS2, and IS3). Candidate selection was performed separately on each tumour’s set of results using all combinations of five different verification budgets (i.e. number of calls selected) and six different selection strategies. F₁ scores were calculated for each set of selected calls and compared to F₁ scores calculated from the full prediction set. To compare the effect of the numbers of algorithms used, datasets were further subset using four different metrics