Table 1.
Area under ROC (TPR vs. FPR) and precision-recall (PPV vs. TPR) curves for off-target scoring methods when benchmarked with the Haeussler dataset [11], allowing up to six mismatches, and NGG, NAG, and NGA PAM sequences for off-targeting
| Off-target scoring method | ||||||||
|---|---|---|---|---|---|---|---|---|
| Area | CRISPRoff | Elevation | CFD | MIT | Cropit | CCTop | VfoldCAS | |
| ROC | .98 | .96 | .96 | .96 | .91 | .88 | .80 | |
| PR | .18 | .08 | .08 | .12 | .05 | .06 | .01 | |
| TPR | ||||||||
| .9 | FPR | .06 | .11 | .11 | .13 | .27 | .34 | .44 |
| .95 | .11 | .17 | .17 | .21 | .33 | .44 | .63 | |
| .99 | .32 | .88 | .88 | .44 | .71 | .74 | .84 | |
| 1 | .73 | .97 | .97 | .96 | .99 | .91 | .96 | |
| FPR | ||||||||
| .01 | TPR | .67 | .52 | .52 | .59 | .36 | .31 | .18 |
| .05 | .89 | .80 | .80 | .79 | .49 | .50 | .39 | |
| .1 | .94 | .89 | .89 | .87 | .71 | .61 | .48 | |
Corresponding TPR and FPR performance of the methods are also given for some fixed FPR and TPR values. Best performances are given in bold