It works, but there is a problem with the results.
In other words, SSM gives you something you would not expect or,
on contrary, does not give you the expected. This is actually not
a rare case. Yes, there may be a bug in SSM that we are not aware
about. Just as well, your expectations may be unjustified. However,
more often the problem roots in that users instruct SSM to do
things that are different of what they, the users, actually want.
We therefore suggest that you pick a category from the following
list which you think best matches your results in order to see
whether your problem is really a problem. And if it is, please
let us know about that.
- SSM says there are no matches while
I know for sure they should be there.
- SSM says my protein has no secondary structure,
while Rasmol shows a beautiful picture of it.
- SSE assignments in SSM are different of what
I got from somewhere else.
- SSM seems to lose certain SSEs
- SSM reports certain SSEs as unmatched
in the "SSE alignment" table, however I see that they are mapped
onto each other in the "3D structural alignment" table just below
- None from the above.
- SSM says there
are no matches while I know for sure they should be
- Yes, theoretically it is possible to find a structural neighbour
for any polypeptide. The only question is how remote that
neighbour would be. When searching the PDB or SCOP archive, SSM
does not pretend to find a PDB or SCOP structure closest to your
input. The task is slightly different: to find all structures
showing a certain measure of similarity to the input. This means
that "less similar" structural neighbours (which make perhaps 90%
of the PDB for a typical query) are not taken into account.
The only reason for taking this approach is to save your time:
searching through the whole PDB is expensive.
Everything is however under your control, because SSM allows you
to specify the desired level of similarity. In the Submission
Form, you are asked to specify the "Lowest acceptible match" for
the query and target structures. By default, the similarity levels
are set at 70%. This means that SSM will consider only matches
where at least 70% of secondary structure elements (SSEs) of the
query may be mapped, or superposed in 3D, on at least 70% of
target's SSEs. By bringing these figures down you can expand
the set of structures to be searched. If both levels are set to
zero, the whole PDB is searched through, and then a match
should be there.
- SSM says my protein
has no secondary structure, while Rasmol shows a beautiful
picture of it.
- Don't be fooled by Rasmol. It first reads the annotation records
in PDB files (HELIX/SHEET), and shows what they say. Remove
these records in your file and load it into Rasmol again. If
secondary structure has not changed, then check for another,
quite frequent, reason:
corrupted (read this link!) PDB file.
If you find that your PDB file complies with the standard,
try to think of what else could hinder the secondary structure
identification. E.g., there was a case when a user submitted
an artificial structure consisting of two chains superposed
on each other. Because of this superposition, the hydrogen
bonding got completely screwed up and SSEs could not be
identified. Intriguingly, in that case SSEs could be found
for each chain separately, but not for the whole protein.
That caused quite a difference in results when the same
structure was submitted as two files, each containing
a single chain.
If you cannot find a reason why SSM keeps saying that there
is no secondary structure motifs in your protein, please
tell us about that.
- SSE assignments in
SSM are different of what I got from somewhere else.
- This is natural if the differences are small, not more than
1, maximum 2, residues on both ends of an SSE. Algorithms
for secondary structure calculations have a few empirical
parameters, and may give different priorities to geometrical
aspects of the structure and its hydrogen bonding. SSM tends
to report longer (by 1-3 residues) SSEs than DSSP. Sometimes
two helices or two strands may merge, as compared to another
SS-calculating algorithm, or, in opposite, an SSE may appear
split in two parts.
If you find that SSE indentification is significantly different
of your expectations, please check first whether your PDB file is
corrupted. If you are certain that SSM is in fault, please
let us know that.
- SSM seems to lose
- This may be due to the general reasons listed
right above. On top of that, SSM ignores
strands shorter than 4 residues and helices shorter than 6
residues, because it was found that too short secondary structure
elements tend to manifest themselves less regularly within
structural families. If this does not explain your observations,
please let us know that.
- SSM reports certain
SSEs as unmatched in the "SSE alignment" table, however I see
that they are mapped onto each other in the "3D structural
alignment" table just below.
- This is normal. SSM compares structures in two steps. In the
first step it matches the protein's secondary structures, and
the results are used as an intial guess for the second step,
that is a precise alignment of Ca's
in 3D. The "SSE alignment" table presents the results of 1st
step. Certain SSEs may not match because of considerable
differences in their positions and orientations relatively
to other, already matched, SSEs. If such SSEs are found on a
considerable distance from each other at best superposition
of the structures, their Ca's do
not get mapped. However, SSEs are also considered as non-matching
if they significantly differ in number of residues. Thus, SSM
does not match a strand of 5 residues to a strand of 10 residues
even if they overlap well at best structure superposition.
However, Ca-alignment would map
individual Ca's of such SSEs.
Although the above may sound somewhat confusing, we decided
to leave the things as is for having an extra indication of
structural dissimilarity. The "SSE alignment" table may show
in a grasp that there are considerable differences in the SSE
topology and/or composition, while the "3D structural alignment"
table delivers further details, allowing for figuring out how
significant the differences actually are.
- None from the above is
- Ok, please help us and we'll do our best to help you:
report your problem.