![[Home]](/icons/homeicon.gif){kind=icon}
Protein Structure Prediction Center
|
DAL_<w> |
DAL_<w>_regions |
|---|---|
| <score> | 2D T M <n_aa1> <n_aa2> ... |
<w> – shift window
<score> – percentage of aligned residues
2D – link to the alignment maps (see explanation below)
T – link to the target structure with aligned regions highlighted
M – link to the model structure with aligned regions highlighted
<n_aak> – size of the k-th region (link points to the corresponding rigid body superposition)
Descriptor-based alignments show similarity of the structures in the non-rigid body mode. As proteins are flexible and undergo significant structural changes in vivo, it is valuable to have a method that allows to reveal structural relationship or detect more comprehensive similarity between the proteins containing topologically similar local substructures. This often can not be detected when the substructures are forced into the same frame of reference of the rigid-body superposition.
Descriptors are local regions of tertiary protein structure. Each encompasses 3D fragments of protein’s backbone and is associated with one specific CA in the protein’s amino acid sequence. As opposed to supersecondary structure defined as a set of secondary structure elements, descriptors deal only with a vicinity of one particular residue. They may be, however, discontinous. Descriptors consist of main chain sections that are 5 residues long, called elements. Each element is centered around either the central residue of the descriptor or around a residue that is in contact with the central residue. Sets of overlapping elements (along the sequence) constitute a segment. As seed descriptors we only consider the descriptors of at least 3 separate segments. Seed descriptors are used for initial alignments of the structures. The original alignment obtained with the seed descriptors can be extended if the two structures exhibit additional similarity in that region.
Descriptors are considered similar, if there exists at least a partial match between their elements which satisfies the following conditions:
central residues are aligned
match is consistent (it yields alignment between residues)
elements around central residues have RMSD no greater than 1.5 Å
other matched elements have RMSD no greater than 2.0 Å
overall RMSD of the aligned residues is no greater than 2.5 Å
at least 75% of residues in both descriptors are aligned
To compare two structures we compare all their descriptors. Each
pair of residues can belong to one or more alignments between
descriptors. These data are searched for maximal alignments.
Such alignments in general do not correspond to a single rigid
body superposittion. We have computed alignments for the following
window sizes:
| Column name | Shift window (residues) |
| DAL_0 | 0 (sequence dependent) |
| DAL_4 | 4 |
| DAL_I | ∞ (sequence independent) |
Note that the program will return a score of 0 if the fold of either structure is relatively simple and therefore it is impossible to build at least one seed descriptor (for example, a structure consisting of two long helices and a turn) or the algorithm can not detect at least one pair of similar seed descriptors in the compared structures.
Color code:
green – segments of the alignment within a given sequence window (segments which are scored)
yellow – other segments of the alignment
cyan – main segments of similar descriptors
purple – other segments of similar descriptors
Borders:
red – α-helix
yellow – β-strand
Target is mapped to vertical axis, model to horizontal. Grid lines are 50 residues apart.
Descriptor alignments are based on local similarity, and thus do not necessarily yield a meaningful rigid-body superposition. Therefore we have computed maximal sets of descriptors with the overall RMSD below 5 Å. Targets are represented with a thick line and models are shown as thin lines in the RasMol representations (backbone).
Color code:
green – segments of the alignment within a sequence window (scored segments)
yellow – other segments belonging to the descriptor alignment
purple – target regions outside of the area of descriptor similarity (if the whole target is aligned then this color won't be seen in the RasMol representation)
red – model regions outside of the area of descriptor similarity (if the whole target is aligned then this color won't be seen in the RasMol representation)
Protein Structure Prediction Center