Table 4.
DeeReCT-PolyA with leave-one-motif-out test on the Dragon human dataset
| Variants | Size | Error Rate (%) |
|
|---|---|---|---|
| 5-fold cross-validation | leave-one-motif-out | ||
| AATAAA | 5190 | 11.81 | 14.20 |
| ATTAAA | 2400 | 9.00 | 8.17 |
| AAAAAG | 1250 | 5.77 | 5.45 |
| AAGAAA | 1230 | 7.76 | 7.52 |
| TATAAA | 880 | 7.69 | 7.18 |
| AATACA | 780 | 10.45 | 8.86 |
| AGTAAA | 690 | 9.55 | 7.46 |
| ACTAAA | 670 | 10.72 | 11.16 |
| GATAAA | 460 | 8.04 | 8.04 |
| CATAAA | 410 | 9.02 | 11.46 |
| AATATA | 410 | 8.78 | 8.54 |
| AATAGA | 370 | 4.59 | 4.05 |
| Average | – | 9.57 | 10.08 |
Note: For the leave-one-motif-out test, for each PAS variant, a DeeReCT-PolyA model was trained with data of all the other motif variants and then test only on this variant. Bold indicates the error rate of best model for each PAS variant.