Table 5.
Results of our clustering-based approach on the human, mouse and rat genome data.
| Human | Mouse (All) | Mouse (Distinct) | Rat | |
| # of predictions | 374 | 381 | 359 | 276 |
| # of TPs | 196 | 172 | 165 | 101 |
| # of real miRNAs | 240 | 245 | 212 | 146 |
| SE | 81.67% | 70.20% | 77.83% | 69.18% |
| Change in SE | -7.91% | -11.02% | -8.09% | -17.81% |
| PPV | 52.41% | 45.14% | 45.96% | 36.59% |
| Change in PPV | +21.25% | +14.81% | +17.37% | +15.94% |
This table shows the SE and PPV obtained after our filtering approach is applied to the predicted candidates generated by ProMirII-g. The PPV is increased by more than 15% in all the three genomes. The SE is kept reasonably high as it is just slightly decreased by a percentage less than 10% in the human and mouse genomes.