Fig. 5

When tested against in silico SARS-CoV-2 variants, QuickVariants demonstrates higher accuracy compared to bcftools by identifying (a) almost no FP indels (0.0–1.0 for 95% CI, median = 0.0), 4.3-fold fewer median FN indels, almost no false positive (FP) point mutations (0.0–1.0 for 95% CI, median = 0.0), and no false negative (FN) point mutations (0.0–3.0 for 95% CI, median = 1.0). a Distribution of FP indels and FN indel rates identified by QuickVariants and bcftools. Variance in FP indels and FN indel rates identified by QuickVariants and bcftools at different mutation frequencies, with datasets labeled by the number of in silico point mutations inserted into the corrected assembly. Distribution of FP point mutations and FN point mutation rates identified by QuickVariants and bcftools. Variance in FP point mutations and FN point mutation rates identified by QuickVariants and bcftools at different mutation frequencies, with datasets labeled by the number of in silico point mutations inserted into the corrected assembly. In both point and bar plots, error bars signify a 95% confidence interval (n = 30 for point plots and n = 240 for bar plots, 8 mutational densities, 10 replicates, and 3 aligners), and the central points denote the mean. For box plots, error bars also indicate a 95% confidence interval, and the central bars represent the median, excluding outliers. b,c Indel identification in sample with 200 in silico point mutations and 20 in silico indels. b Illustration of three examples where bcftools (indel) excludes the insertion for read 2. A “-” represents a deletion. c FN indels identified by QuickVariants and bcftools. Both QuickVariants and bcftools identified 0 FP indels in this sample