Skip to main content
. 2021 Sep 24;10(9):giab063. doi: 10.1093/gigascience/giab063

Table 2:

Benchmarking SV recall, precision, and F1 on HG002 Human (hg19) PacBio reads (CLR and HiFi) at varying coverages (CLR 10×, 20×, 30×; HiFi 10×)

Method Recall, % Precision, % F1, %
PacBio CLR 10×
minimap2 62.85 88.88 73.63
NGMLR 60.11 86.44 70.91
Vulcan
 60% 60.13 89.41 71.90
 70% 60.79 90.12 72.61
 80% 60.97 90.13 72.73
 90% 61.85 89.93 73.29
PacBio CLR 20×
minimap2 77.76 71.85 74.69
NGMLR 75.74 68.36 71.86
Vulcan
 60% 75.74 74.69 75.21
 70% 75.98 75.32 75.65
 80% 76.22 75.65 75.93
 90% 76.90 75.08 75.98
PacBio CLR 30×
minimap2 83.71 86.25 84.96
NGMLR 81.79 82.41 82.10
Vulcan
 60% 82.05 86.31 84.12
 70% 82.33 87.12 84.66
 80% 82.47 87.60 84.96
 90% 82.75 87.49 85.05
PacBio HiFi 10×
minimap2 81.50 90.70 85.85
NGMLR 78.22 86.26 82.04
Vulcan
 60% 77.73 86.04 81.68
 70% 77.74 86.19 81.75
 80% 76.40 85.75 80.81
 90% 76.26 85.73 80.72

Various percentile cut-offs for Vulcan were used, including 60%, 70%, 80%, 90%. Vulcan achieves the highest F1 score on PacBio CLR 20× and 30× reads, with minimap2 achieving the highest F1 score on PacBio CLR 10× and PacBio HiFi 10× reads.