Table 4.
Computing Time for Different Modules
| Module | # Selected atomsa | Approximate CPU timeb |
|---|---|---|
| Surface (WordomARVO)c | 115d | 2980 |
| Surface (ARVO)e | 115 | 3690 |
| Surface (WordomGEPOL–ASURF)f | 115 | 2130 |
| Surface (GEPOLASURF)g | 115 | 2660 |
| Surface (WordomGEPOL–ESURF)h | 115 | 5900 |
| Surface (GEPOLESURF)i | 115 | 7290 |
| Surface (WordomGEPOL–WSURF)j | 115 | 1890 |
| Surface (GEPOLWSURF)k | 115 | 1970 |
| Correlation (DCC)i | 360m | 4 |
| Correlation (LMI)n | 360 | 63 |
| PSNo | 2593 | 391 |
| PSN-path | – | 15 per pair |
| Clustering (distances only)p | 316q | 1461 |
| Clustering (QT-like)r | 316 | 100 |
| Clustering (hiero)s | 316 | >50,000 |
| Clustering (leader)t | 316 | 10 |
| Clustering (leader)u | 316 | 10 |
| Clustering (leader)v | 316 | 45 |
The considered system is a 10,000 frame trajectory of the GTP-bound Gαi1 subunit (PDB: 1CIP; 2593 atoms; 316 residues and 1 GTP molecule (44 atoms)).
CPU time (seconds) on an AMD Athlon 64 3000+, 2 GHz, 2 GB RAM.
Solvent accessible surface area computed by the Wordom implementation of the ARVO algorithm.
selection consisted in GTP and first 9 residues (selection /*/@(1–10)/*)
Solvent accessible surface area computed by the ARVO program.
Solvent accessible surface area computed by the Wordom implementation of the GEPOL algorithm (highest accuracy).
Solvent accessible surface area computed by the GEPOL program (highest accuracy).
Solvent excluded surface area computed by the Wordom implementation of the GEPOL algorithm; accuracy settings: rmin 0.5, ofac 0.8, ndiv 5.
Solvent excluded surface area computed by the GEPOL program; accuracy setting: rmin 0.5, ofac 0.8, ndiv 5.
van der Waals surface area computed by the Wordom implementation of the GEPOL algorithm; highest accuracy.
van der Waals surface area computed by the GEPOL program; highest accuracy.
Residue-residue correlation by means of the dynamic cross correlation method; masses were not taken into account.
Selection consisted in all Cα atoms and GTP
Residue-residue correlation by means of the linear mutual information method; masses were not taken into account.
PSN analysis probing 11 different Imin values (from 0.0 to 5.0 with a 0.5 step).
Only the RMSD-based distance matrix was computed at this stage and written to file.
All Cα atoms were selected.
Clustering by the QT-like algorithm, using a precalculated distance matrix (RMSD cutoff 1.0 Å).
Clustering by the hierarchical algorithm, using a pre-calculated distance matrix (RMSD cutoff 1.0 Å).
Clustering by the leader-like algorithm (RMSD cutoff 1.0 Å); distance matrix is not necessary.
Clustering by the leader-like algorithm (RMSD cutoff 1.0 Å) and turning on the non-markovian option. In this case, the bottleneck is disk speed (CPU usage 18%).
Clustering by the leader-like algorithm (DRMS cutoff 1.0 Å).