aMethod: | I | II | III | IV | V |
|---|
T | 6854696 | 6854696 | 6854696 | 6854696 | 6854696 |
TP | 2101562 | 3603662 | 5489738 | 2531549 | 5036243 |
FN | 4753134 | 3251034 | 1364958 | 4323147 | 1818453 |
F | 5919255 | 5919255 | 5919255 | 5919255 | 5919255 |
FP | 20437 | 220957 | 1085303 | 9319 | 46906 |
TN | 5898818 | 5698298 | 4833952 | 5909936 | 5872349 |
bMethod: | I | II | III | IV | V |
Sn | 30.66% | 52.57% |
80.09%
| 36.93% | 73.47% |
Sp | 99.65% | 96.27% | 81.66% |
99.84%
| 99.21% |
Acc | 62.63% | 72.82% | 80.82% | 66.08% |
85.40%
|
Mcc | 99.04% | 94.22% | 83.49% |
99.63%
| 99.08% |
Ppv | 30.57% | 50.93% | 69.14% | 36.88% |
72.97%
|
Pc | 40.61% | 53.18% | 61.61% | 45.94% |
74.04%
|
F1 | 46.82% | 67.49% | 81.75% | 53.89% |
84.37%
|
- I:CpGPlot, II:CpGReport, III:CpGProd, IV:CpGCluster, V:GaussianCpG
- For Panel a: The unit of measurement is necleotide
- True, T: the length of known CpG islands
- False, F: the length of non-CpG islands
- True positive, TP: the length of predicted known CGIs
- False positive, FP: the length of predicted CGIs not in known CGIs
- False negative, FN: the length of not predicted known CGIs
- True negative, TN: the length of predicted non-CGIs
- For Panel b:
- Sensitivity, Sn=TP/(TP+FN)
- Specificity, Sp=TN/(TN+FP)
- Accuracy, Acc=(TP+TN)/(TP+FP+FN+TN)
- Mean correlation coefficient,
-
\({\qquad }{Mcc}={\textstyle {{{TP}\times {TN} - {FN}\times {FP}} \over {\sqrt {({TP} + {FN}) \times ({TN} + {FP}) \times ({TP} + {FP}) \times ({TN} + {FN})} }}}\)
- Positive predictive value, Ppv=TP/(TP+FP)
- Performance coefficient, Pc=TP/(TP+FN+FP)
- F1 score, the harmonic mean of precision and sensitivity,
-
F1=2×TP/(2×TP+FP+FN)
- For Panel a&b: Default parameters for all software are set
