| 40X | 30X | 20X | 10X | 4X |
---|
Method | SE | SP | ABE | SE | SP | ABE | SE | SP | ABE | SE | SP | ABE | SE | SP | ABE |
---|
B-phon | 12/12 | 12/12 | 0.96 | 12/12 | 12/12 | 1.0 | 12/12 | 12/12 | 0.96 | 11/12 | 11/11 | 1.36 | 5/12 | 5/5 | 0.5 |
SVD | 11/12 | 11/11 | 2.7-417 | 11/12 | 11/11 | 4.8-405 | 11/12 | 11/11 | 7.4-389 | 11/12 | 11/11 | 14.7-370 | 8/12 | 8/9 | 29-336 |
BD | 12/12 | 16/16 | 185.3 | 12/12 | 16/16 | 189.1 | 11/12 | 14/14 | 171.3 | 10/12 | 15/15 | 137.6 | 4/12 | 4/4 | 153 |
GASV | 12/12 | 12/12 | 100-226 | 12/12 | 12/12 | 103-245 | 12/12 | 12/12 | 106-257 | 12/12 | 12/12 | 114-270 | 10/12 | 10/10 | 129-315 |
CREST | 12/12 | 25/25 | 1.2 | 12/12 | 23/23 | 1.2 | 11/12 | 15/15 | 0.9 | 9/12 | 11/11 | 0.8 | 5/12 | 5/5 | 1.1 |
- Sensitivity, specificity, and breakpoint error of interchromosomal breakpoint calls across all five coverage levels for all five programs. SE = sensitivity (true events captured/ # true events), SP = specificity (# predictions that capture true events/ # predictions), ABE = average breakpoint error. Since SVDetect and GASV predict a range of breakpoints, we reported their average breakpoint errors as ranges. BreakDancer and CREST tended to have redundant predictions, but we did not count this against specificity since redundant predictions typically supported a single common event. These redundancies could easily be merged into a single prediction. The CREST package includes a program to remove redundancies, but BreakDancer does not address these redundant predictions.