Table 6 By randomly choosing 10 subsets from each of the 5 datasets whose positive subset departs from the negative one greatly, we try to alleviate the bias between the value of sensitivity and specificity. Here we show the average results tested on Satron's dataset with cut-off value as 0.6. All the sub-datasets and results have been supplied in the additional file (see additional file 1 and additional file 3).
From: Demonstration of two novel methods for predicting functional siRNA efficiency
| Â | A+B+C | A+B | B+C | A+C | A | B | C |
|---|---|---|---|---|---|---|---|
Accuracy | 68.43 ± 1.45% | 67.13 ± 1.95% | 65.37 ± 2.19% | 68.09 ± 2.14% | 68.76 ± 2.59% | 61.55 ± 2.56% | 63.90 ± 2.52% |
Sensitivity | 69.05 ± 1.20% | 69.78 ± 2.98% | 64.33 ± 3.41% | 67.25 ± 3.19% | 69.72 ± 3.63% | 63.93 ± 2.96% | 58.03 ± 3.63% |
Specificity | 67.81 ± 3.07% | 64.49 ± 2.21% | 66.4 ± 3.36% | 68.93 ± 2.53% | 67.81 ± 2.68% | 59.16 ± 3.63% | 69.78 ± 4.32% |
Pearson | 0.4683 ± 0.03219 | 0.4618 ± 0.03205 | 0.4112 ± 0.03306 | 0.4447 ± 0.03097 | 0.4718 ± 0.03558 | 0.3758 ± 0.02486 | 0.3196 ± 0.02920 |
ROC | 0.7452 ± 0.02383 | 0.7255 ± 0.02886 | 0.7093 ± 0.02450 | 0.7367 ± 0.02454 | 0.7411 ± 0.02933 | 0.6768 ± 0.02220 | 0.6715 ± 0.02659 |