TABLE 2.
Accuracy on randomly sampled short subsequences in terms of ND from the original position
| Data | EPA | EPA-RA | BLAST |
| Outer QS | |||
| D140 | 0.49 | 0.58 | 0.82 |
| D150 | 1.14 | 1.2 | 1.96 |
| D218 | 1.7 | 2.01 | 3.66 |
| D500 | 1.31 | 1.37 | 2.36 |
| D628 | 2.44 | 2.95 | 2.69 |
| D714 | 1.71 | 1.82 | 2.36 |
| D855 | 2.87 | 2.97 | 3.53 |
| D1604 | 2.26 | 2.45 | 2.87 |
| Inner QS | |||
| D140 | 0.74 | 0.71 | 1.81 |
| D150 | 3.09 | 3.1 | 4.62 |
| D218 | 2.24 | 2.66 | 3.81 |
| D500 | 2.05 | 2.37 | 4.16 |
| D628 | 3.28 | 3.65 | 4.04 |
| D714 | 1.78 | 1.93 | 3.51 |
| D855 | 3.68 | 3.74 | 4.91 |
| D1604 | 2.23 | 2.38 | 3.86 |
Notes: The second column (EPA) shows the average ND of the EPA placements (using slow insertions under the GTR + Γ model) for the data sets in question. The third column (EPA-RA) shows the average ND for the EPA with previous realignment using HMMER. The last column (BLAST) shows the average ND for a BLAST-based approach.