Fig. 1

150-bp ROCker models for A aioA, B arxA, C arrA, and D arsC2 and comparison of ROCker models versus different fixed cut-offs: id70, id80, and id90. Three parts of the ROCker plots show: (TOP) Positive reference sequence alignments with amino acids in different colors and gaps in light grey are displayed in the top panel. (Middle) Bit scores (y-axis) of the hits (i.e., the matches from simulated 150-bp read datasets) from positive references (blue), negative references (yellow), and nontarget sequences (red). The solid black traversing line represents the calculated ROCker best bit score thresholds for consecutive windows of variable length. (Bottom) Summary statistics on the performance of each window are based on true/false positives (TP and FP, respectively) and true/false negatives (TN and FN, respectively). The accuracy (i.e., [TP + TN]/[TP + FP + TN + FN]) of ArxA and AioA was 99.99%, that of ArsC2 was 99.61%, and that of ArrA was 99.97%. Boxplot components: centerline, median values; box limits, upper and lower quartiles; whiskers, 1.5 × interquartile range; points, outliers