Help and troubleshooting
The web interface uses an algorithm to accurately compute flexible regions of a protein structure, given two different conformations of a protein.
Specifying protein structures
You can specify protein structures either by their PDB code, or upload PDB-formatted files yourself (e.g. to measure flexibility in user-created structures). For a structure, you have the option to specify a chain or a range of residues.
Note:
- Nucleic acids and other molecule types are not currently supported. Modified amino acids are permitted provided that each residue has backbone atoms named "N", "CA", and "C".
- For ensembles of structures, pFlexAna takes only the first model.
- Crystallographic symmetries are not reconstructed.
- Ligands and solvent molecules are automatically ignored.
Specifying input parameters
The parameters required are:
- σ, noise level for coordinates
: A lower value of σ means that two fragments will need to be
more similar
(e.g. have lower RMSD) to be considered rigid. We recommend values
from 0.2 to 0.8, and suggest that lower values (stricter similarity
checks) be used for highly homologous proteins. In practice, we have
found values of upto 0.4 work well for reasonably close homologs (40% identity
or higher). Higher values of σ may be
used to detect weak similarities between
evolutionarily distant proteins. We use a statistical test to determine the
structural similarity of a pair of fragments, and model coordinate noise using
a Gaussian model. σ corresponds to the standard deviation of our model.
- k, the number of desired clusters: The final output of the algorithm (see below) is a list of aligned fragment-pairs. These fragment pairs will be clustered into k clusters such that the dominant conformational motion occurs between clusters. k corresponds to the number of rigid units (domains, lobes, etc.) that the user expects the protein to have. We recommend a starting with k = 2, and modifying this parameter in later runs if needed.
Please go to the input page to run our algorithm.
Output
A sample output page can be seen here.
The first table shows the aligned fragment pairs that are structurally similar. These are clustered into k clusters.
For each aligned fragment pair, we show below it's transformation, RMSD and a cartoon picture highlighting the area of similarity, which aids in visualization of any motions that are relative to it.
Problems?
Please e-mail us at anshulni -at- comp dot nus dot edu dot sg to report a problem.