On 2/3/10 5:52 PM, Thomas Evangelidis wrote: > I have a protein domain with very low sequence identity to PDB > structures (37%), therefore I created a nice consensus alignment with > T-COFFEE which includes both sequences and templates. After carefully > inspecting the alignment, discarding bad matches, and realigning, I > ended up with 8 potential templates which all look good to me. Thus I > tried to adapt the compare.py script used in the basic tutorial, to help > me select the best templates. ...
You don't have to use Modeller to generate your alignment - you can generate it however you like (e.g. manually, with your favorite program, etc.). There is also no requirement to align your templates with malign3d (as you noted, that won't work unless the templates have a lot of structural overlap anyway). So if you like your T-COFFEE alignment, by all means use that directly (subject only to modifying the file format so that Modeller can read it).
Ben Webb, Modeller Caretaker