Set | FATCAT residue pairs | FIRE performance | T-Coffee performance | ClustalW performance | MAFFT performance |
|---|
1
| 54 | 0.87 | 1.00 | 0.99 | 1.00 |
2
| 94 | 0.57 | 0.96 | 0.84 | 0.97 |
3
| 137 | 0.83 | 0.97 | 0.83 | 0.96 |
*4
| - | 0.69 | 1.00 | 0.68 | 0.94 |
5
| 103 | 0.83 | 0.98 | 0.71 | 0.88 |
6
| 108 | 0.87 | 0.97 | 0.73 | 0.87 |
- Performances of the FIRE, T-Coffee, ClustalW and MAFFT algorithms were measured by determining the proportion of correctly aligned residue pairs using FATCAT and DALI structure-based alignments as a reference. FIRE is independent of homology and performed better than ClustalW for data sets 5 and 6 (antibody variable regions), illustrating the value of using this approach when sequence similarities are low. This independence from residues in the sequence may also lead to relatively poor FIRE performance when sequence similarity is high, for example set 2. T-Coffee and MAFFT performed best overall (although see text for further discussion). The PDB structure files included in FATCAT, DALI and T-Coffee algorithms are set 1: 2DIM, 2K9N; set 2: 2DIM, 2DIN; set 3: 2YUM, 2K9N; set 4: 1BO5; set 5: 5LVE, 1QP1; set 6: 1LVE, 1NC4. *Due to a lack of structural data for set 4, the FUGUE threading algorithm [23] was used to generate a reference structure alignment from the E. histolytica sequence (XM_650121.1) using E. coli glycerol kinase (PDB IB: 1BO5) as a template. For all alignments, FATCAT and DALI produced similar results and only the FATCAT data are shown.
