Table 2.
Performance comparison of SRAMP and BERMP on the independent mammalian dataset at various stringency thresholds.
| Mode | Stringency (Specificity) | SRAMP | BERMP | ||
|---|---|---|---|---|---|
| Sensitivity | MCC | Sensitivity | MCC | ||
| Full transcript mode | Very high (98.7%) | 25.7% | 0.373 | 29.6% | 0.421 |
| High (93.7%) | 50.3% | 0.414 | 60.3% | 0.492 | |
| Moderate (88.1%) | 64.5% | 0.405 | 74.9% | 0.475 | |
| Low (83.0%) | 72.8% | 0.385 | 82.5% | 0.447 | |
| Mature mRNA mode | Very high (99.1%) | 11.0% | 0.211 | 11.0% | 0.215 |
| High (95.0%) | 29.6% | 0.273 | 33.5% | 0.309 | |
| Moderate (90.0%) | 44.0% | 0.293 | 48.7% | 0.325 | |
| Low (85.3%) | 54.2% | 0.294 | 58.9% | 0.325 | |
Note: The very high, high, moderate and low stringency thresholds correspond to approximately 99%, 95%, 90% and 85% specificities in five-fold cross-validation tests, respectively. The same datasets were used to develop and compare both classifiers (Figure S1). The results for SRAMP excerpted from 15.