Jairo Rocha wrote: > I have some sequences from real bacteria that correspond to one protein > in the pdb, 1SIG. > Of course, they have lots of mutations. I would like to know which > changes the structure would have, due to these mutations. > > I used Modeller to answer that question, for the first time. The > predicted structures are very very similar to the given structure. > I am surprised that some helices are not broken or bended just a bit. ... > As far as I know, to get a helix requires several energy states to > coincide, so it is difficult that by change I get the helix. > So it seems to me the program just copies the structure.
Modeller is a package for comparative modeling, so by construction your target model is going to look like the template(s). It doesn't exactly "copy" the structure, but certain structural features (e.g. dihedral angles, CA-CA distances, etc.) in aligned templates will be preserved in the model. So if you have a template aligned that has helical structure, it is practically certain that the model will generate will also be helical, regardless of the actual sequence. (Of course, something like poly-P probably wouldn't be helical, but then that's probably a failure of your alignment method if you've aligned poly-P with a helical sequence anyway.)
You probably don't want homology information if you're investigating mutations with structural changes. You can certainly use Modeller for this though - just don't use the homology information (see the mutate model script in the Modeller wiki for an example).
Ben Webb, Modeller Caretaker