Table 4.
Classification performance for 3 sets of soluble or transmembrane proteins naively tested. "Observed as" corresponds to the number of residues observed at a TM or a non-TM position. "Predicted as" corresponds to the number of residues predicted at a IPM or non-IPM position. "Proteins with TM α-helix" is a set of 101 proteins with 1 or more TM α-helices. "Proteins with TM β-barrel" is a set of 21 TM β-barrel proteins. TM proteins are extracted from the MPtopo database (3D_helix and 3D_other subsets, respectively). "Soluble proteins" is a set of 65 soluble proteins extracted from the PDB (sequence similarity < 25%). These 3 sets were submitted to the sequence-to-topology SVM, using PHAT and a positional weighting (Table 1). An average prediction was then computed for each sequence of the sets following the procedure described above (Table 3).
| Proteins with TM α-helix | Proteins with TM β-barrel | Soluble proteins | ||||
| Observed as | Observed as | Observed as | ||||
| Predicted as | TM | non-TM | TM | non-TM | TM | non-TM |
| IPM | 181 | 152 | 16 | 5 | - | 57 |
| non-IPM | 11057 | 14423 | 3540 | 4138 | - | 30310 |
| Total number of residues | 11238 | 14575 | 3556 | 4143 | - | 30367 |