OK, figured it out with help from the suggestions. Thanks !
PROTSKIN, given a PDB file and a CLUSTALW alignment does a fine job of
modifying the B-factor column into % identity (or similarity)
Two problems I encountered:
1. A residue we mutated not matching up with the corresponding sequence
- I mutated the residue back to correspond with the sequence (buried
anyway..)
2. Gaps in the coordinates (eg. omitted flexible loops) cause the
program to choke as the register doesn't match up with the corresponding
sequence in the alignment
- I added "dummy atoms" (Charlie Bond) for the missing residues
corresponding to the missing sequence in the structure.
I used GRASP for the visualisation (instructions on the webpage) with
the property file written out by PROTSKIN. Be careful to remove the
%identities corresponding to the inserted dummy atoms or the assignment
(after the gap) will be wrong !
Thanks again to everyone for their help and suggestions.
best,
Iain
Christophe Deprez wrote:
Hi Iain, thanks for your interest in ProtSkin.
ProtSkin expects the atom sequence in your PDB file to be exactly the
same as the "Query" sequence in your Blast text file.
If this is not the case, you will have to modify one or the other file
to make the sequences match.
There may be different reasons why your alignement does not use the
same sequence as your structure file (which are not always clear to me
I must admit).
In the case of a mutation, if you don't want to run Blast on your
exact PDB sequence and would rather use a given sequence alignement
file, the approach suggested by Charlie should work. Alternatively,
you could also alter the Blast text file (in the "Query" sequence) to
display the same sequence as your PDB file, although this will affect
the number/color for the mutated residue.
In the case of missing residues in the coordinates, I assume you don't
want to compute any alignment for these residues as they will not even
appear in your structure! Use the second approach here by removing the
missing residues from the alignment file (in all aligned sequences).
Please not that the "plain sequence alignment" given on the error page
can be downloaded, altered and submitted again as input file (select
"Sequence alignment from a previous query") in the ProtSkin form.
Please give it a try and let us know.
Christophe Deprez
Iain Kerr wrote:
Of course, should have tried that BUT, it appears PROTSKIN has
problems everytime it encounters a gap. It now complains about res.
296 in the same fashion (287-295 are missing from the coordinates....)
There must be a way around this...I could split all the different
parts (ie. those separated by gaps) into different files but I expect
the smaller parts will align poorly with the sequences or I'll get a
similar error again....
( P.S. sorry !, posted to COOT again by accident, have fixed my
address book..)
Charlie Bond wrote:
Iain, I've never used protskin, but can you not just mutate the
offending residue in your PDB file to get protskin to run (eg in
coot or pdb-mode). Then afterwards you can manually alter the
B-factor for that residue to what it should actually be and then run
grasp etc.
Cheers,
Charlie
Iain Kerr wrote:
I tried that. My protein has a mutation at res. 141, so it said:
*Error at residue 141 :
The sequence in your PDB file does not match your BLAST query
sequence.
*Here is the plain sequence alignment
<http://www.mcgnmr.ca/ProtSkin/tmp/01232-sequence_alignment>
derived from it.
I deleted res. 141 and re-ran but:
*Error at residue 142 :
The sequence in your PDB file does not match your BLAST query
sequence.
*Here is the plain sequence alignment
<http://www.mcgnmr.ca/ProtSkin/tmp/01232-sequence_alignment>
derived from it.
and 142 is correctly assigned !
This looked like it was going to work !
Iain
Mark Brooks wrote:
Hi,
did you try Protskin? (http://www.mcgnmr.ca/ProtSkin/)
Its a lot less hassle than Espript, and makes Grasp input files as
well as for Pymol
Good luck,
Mark
On 31/01/07, Iain Kerr <[EMAIL PROTECTED]> wrote:
I'm trying to colour a molecular surface by sequence
conservation...(sorry,
I think I incorrectly posted this to COOTBB the other day)
I've figured out how to do it in GRASP - modify the B-factor
column in the
PDB file to represent the percentage conservation and then colour
the
surface by B-factor. I know ESPRIPT will make the modified file,
but I'm
having trouble generating the correct one..
I am providing ESPRIPT (expert mode, %Equivalent' scoring
function) with a
CLUSTALW alignment (Aligned sequences > Main alignment file) and
a PDB file
('Aligned sequences' > 'Supplementary pdb' file). I get the error:
'Fatal error: wrong format in PDB file.'
..and the values in the B-factor (%Equivalent) column are all
either 99 or
100 which is nonsense according to the alignment.
Has anyone come across this. I don't see anything wrong with my
PDB file..
Thanks,
Iain
