I recently used an ensemble of many proteins with Phaser. A single model did not find a solution but this ensemble of seven very similar molecules gave a correct solution with a TFZ of ~9.0. I did not trim the side chains in this ensemble (first try worked) but this introduced variation may be beneficial in the search.
Also try FFAS03 to identify the closest models (not just sequence based) http://ffas.ljcrf.edu/ffas-cgi/cgi/ffas.pl
There are instructions in phaser manual (shown below) about how to generate a single model based on these proteins.
-------------------------------------------------------------------------------------------------------------------------------------------
How can I obtain a "mixed model", as described by Schwarzenbacher et al. (Acta Cryst. D60:1229-1236, 2004)?
Use the FFAS server maintained by the Godzik lab. However, the procedure to get a "mixed model" with the original conformation of conserved side chains is not immediately obvious.
* Enter your sequence, choose the PDB database, click Search, and when the search is finished you'll be at the Results page.
* From that page, click on the link to the PDB database in the "Results vs" column.
* For each model of interest, make a note of the percent sequence identity (quite far to the right). You'll need this to give Phaser an idea of the expected RMS error of the model. Then click on the "scwrl" link under "Psi-Blast/align/model". This will open a page on the SCWRL server.
* Click the check box for "Retain original conformations of conserved residues", then click "Submit query".
* Finally, when the result from this appears, right-click "Get mixed model" and choose "Save Link As..." (or whatever the equivalent is on your browser, e.g. click-hold on a Mac) to save the PDB file with the mixed model.
---------------------------------------------------------------------------------------------------------------------------------------------
Also I agree with Jürgen, Balbes can work great and very quickly.
HTH,
Gordon
M. Gordon Joyce, Ph.D.
Intramural AIDS Research Fellow,
Structural Immunology Section,
Laboratory of Immunogenetics,
NIAID/NIH
Twinbrook II, Rm 108,
12441 Parklawn Drive,
Rockville,
MD 20852, USA
Office Phone: 301 594 0242
Lab Phone: 301 496 3792
---------------------------------------------------------------------------------------------------
-----Original Message-----
From: Miri Hirshberg [mailto:[log in to unmask]]
Sent: Sunday, May 23, 2010 3:35 PM
To: [log in to unmask]
Subject: Re: [ccp4bb] Finding best model for molecular replacement
Sun., May 23rd 2010
EBI
Paul,
1. yes you can run your sequence against all PDB.
1. http://www.ebi.ac.uk/pdbe-srv/view/
drop the one letter sequence in the sequence box and search
2. http://www.ebi.ac.uk/Tools/fasta33/index.html
From the Databases protein you pick
Protein structure Sequence
You drop your 1-letter code sequence in the sequence box and search You
both 1&2 run the same search but the layout of the output is a bit different.
3. You can also try http://meta.bioinfo.pl/submit_wizard.pl
it will predict 3D structure using many methods.
Miri
On Sun, 23 May 2010, Paul Lindblom wrote:
>
> Hi everybody,
>
> I just crystallized a new project protein. How can I find a possible
> model for using molecular replacement? I have the sequence of my
> protein. Is it enough to make a sequence search in the pdb? Or is there another approach I can use?
>
> Thanks a lot,
>
> Paul
>
>
>
------------------------------------
Dr Miri Hirshberg
European Bioinformatics Institute UK
PDBe - EBI -EMBL
http://www.ebi.ac.uk/pdbe
Phone: +44 (0) 1223 492647
FAX: +44 (0) 1223 494468
------------------------------------
|