Table 1.
Structure prediction of membrane proteins.
| No. of TMHs/no. of residues | Constraint type | Highest maxsub in 5,000 simulations (old) | Highest maxsub in 5,000 simulations (new) | Enrichment in high maxsub models (new versus old) | Highest maxsub (full/TMH) | Highest maxsub among 5 lowest energy models (full/TMH) | |
|---|---|---|---|---|---|---|---|
| Bacteriorhodopsin subdomain | 4/123 | One predicted | 1.0 | 1.0 | 0.1 | 1.0/1.0 | 1.0/1.0 |
| V-type ATPase subunit | 4/145 | One predicted | 1.0 | 1.0 | 0.5 | 1.0/1.0 | 0.99/1.0 |
| Cytochrome c | 5/191 | One predicted | 0.70 | 0.88 | 40.3 | 0.93/0.96 | 0.91/1.0 |
| Lac permease Nterminal subunit | 6/190 | One predicted | 0.71 | 0.65 | 0.1 | 0.74/0.82 | — |
| Lac permease Cterminal subunit | 6/185 | One predicted | 0.82 | 0.91 | 1.4 | 0.98/0.99 | — |
| Bacteriorhodopsin | 7/227 | One predicted | 0.68 | 0.83 | 8.7 | 0.93/0.96 | 0.89/0.99 |
| Sensory rhodopsin | 7/217 | One predicted | 0.71 | 0.81 | 3.2 | 0.93/0.95 | 0.52/0.54 |
| Halorhodopsin | 7/239 | One predicted | 0.64 | 0.73 | 4.5 | 0.79/0.89 | 0.48/0.57 |
| Bovine rhodopsin | 7/278 | One predicted | 0.56 | 0.52 | 0.2 | 0.58/0.69 | 0.40/0.53 |
| Beta2 adrenergic receptor | 7/282 | One predicted | 0.46 | 0.55 | 6.3 | 0.61/0.74 | 0.36/0.46 |
| Fumarate reductase | 5/216 | Two pairs of histidine binding hemes | 0.59 | 0.73 | 3.2 | 0.86/0.92 | 0.84/1.0 |
| Cytochrome bc1 | 5/222 | Two pairs of histidine binding hemes | 0.43 | 0.62 | 197.2 | 0.66/0.82 | 0.55/0.80 |
| Lac permease Cterminal subunit | 6/185 | One experimentally-determined salt bridge | 0.82 | 0.99 | 47.6 | 1.0/1.0 | — |
The highest maxsub in 5,000 simulations generated without constraints [using the ″old″ version of RosettaMembrane (5)] and with constraints (“new”, current version of RosettaMembrane) is reported in columns 4 and 5, respectively. Maxsub is the fraction of residues in the model that are superimposable within 4 Å to the X-ray structure of the target (18). The increase in the frequency of models with maxsub greater than or equal to that in column 4 in the constrained runs is reported in column 6. The model closest to the native structure generated by the current version of RosettaMembrane using the protocol described in Materials and Methods is reported in column 7 with maxsub given for both the full-length (″full″) and the transmembrane helical core regions [″TMH″, as predicted by Octopus (15)]. The most accurate (highest maxsub) model among the 5 lowest all-atom energy models is reported in the last column for each target except the isolated subunits of the Lac permease.