False-Negative and False-Positive Percentages for (Normalized) Log-Odds Scores From Markov Models, for Various Orders and Alphabets
| Alphabet | Order | |||||||
| 1 | 2 | 3 | 5 | |||||
| FN | FP | FN | FP | FN | FP | FN | FP | |
| 2 | 20.4 | 13 | 19.4 | 9.5 | 24.7 | 5.0 | 14.0 | 12.5 |
| 3 | 23.7 | 9.5 | 11.8 | 18.5 | 16.1 | 10.5 | 9.7 | 6.0 |
| 4 | 21.5 | 11.0 | 19.4 | 9.0 | 12.9 | 10.0 | 1.1 | 3.0 |
| 5a | 9.7 | 12.5 | 9.7 | 7.0 | 10.8 | 4.0 | 0 | 0 |
| 5b | 23.7 | 6.5 | 21.5 | 5.0 | 8.6 | 7.0 | 0 | 0 |
| 7a | 11.8 | 12.5 | 8.6 | 8.0 | 4.3 | 3.5 | 0 | 0 |
| 7b | 16.1 | 7.0 | 8.6 | 6.0 | 3.2 | 0.5 | 0 | 0 |
-
Percentages are obtained as for ASPC in Figure 1A. Alphabets considered here are: (2) match, other; (3) match, mismatch, gap; (4) match, transition, transversion, gap, (5a) match (A orT), match (C or G), transition, transversion, gap; (5b) match (A or G), match (C or T), transition, transversion, gap; (7a) match (A or T), match (C or G), transition, transversion (A–T), transversion (C–G), transversion (other), gap; (7b) match (A), match (C), match (G), match (T), transition, transversion, gap.
-
Fifth order models on alphabets of 5 or more symbols give scores for which regulatory elements and ancestral repeats distributions do not intersect (any threshold between the maximum score value for ancestral repeats and the minimum score value for regulatory elements guarantees 0% false positive and 0% false negatives).
