Jake Gunn-Glanville wrote: > I have been modeling a protein with a template I am quite confident in, > and a loop that I would like to use a separate template for. When I > generate 200 models, they all seem to show a great deal of structural > diversity. Curious, I developed 20 models of a solved structure against > itself, and found a bit of diversity in all of these resulting models as > well. Is there a good way to constrain automodel so that it diverses > less from the templates when thre are no obvious residue clashes forcing > it to make adjustments?

The objective function optimized by Modeller does not encode all of the information in the template (e.g. it also incorporates statistical information such as Ramachandran-like maps) so you will always see some divergence, even with 100% identical templates. Another thing you should be aware of is that the starting structures are not exactly the templates - the coordinates are randomized slightly. You can reduce this randomization for the entire model with the deviation parameter when you create your automodel object (see http://salilab.org/modeller/manual/node42.html); you can dial this down from the default 4A. If you only want to randomize part of the model, you can write your own function to do that and assign it to automodel.rand_method (see http://salilab.org/modeller/manual/node46.html)

Ben Webb, Modeller Caretaker