Fig. 4

smCounter performance on calling 1% SNVs in N0030, based on reduced barcode depth and rpb. For better visualization, only a subset of barcode depth and rpb combinations are plotted. Cutoff selection and detection limit estimation are also based on the same downsampled data. a ROC curves at barcode depth of 3,612 (full data), 2,167, and 722 with rpb fixed at 8.6. b ROC curves at rpb of 8.6 (full data), 4.0, and 1.5 with barcode depth fixed at 3,612. c Optimal cutoffs that give the maximum sensitivity while allowing no more than 20 FP/Mb at varying barcode depth (x axis) and rpb (represented by different symbols). The dashed line is the linear regression equation of cutoff versus barcode depth across all rpb values. d Estimated detection limit as a function of locus-specific barcode depth, overall average barcode depth, and false positive rate (FPR) allowed. The y-axis is the lowest variant allele fraction that can be detected with at least 95% probability based on a binomial distribution