Jean-Paul Armache wrote: > I would like to pose a question. I am doing homology modeling of various > proteins, some of which are > really straightforward to model, others, due to some issues, not. > I have a problem in this case: a protein has one or several folded > domains and long, unstructured parts that > are connecting them. Sometimes on those unstructured parts are very > small folded parts, sometimes not. > What is critical for me is to model these unstructured parts as close to > the original template as possible. > Instead, I land up with models with these loops that are going all ways > but the way it was in the template. > This happens also in the situation when two sequences have really high > identities and similar lengths. > Is there a way to tell Modeller to model these parts as close to the > model as possible?

Modeller will, by construction, build models that look like the template. This should particularly be true where you have high sequence identity. So if the models deviate greatly from the template, either you have a very poor alignment or your model is conflicted in some way. In the latter case you should see this from the reported violations.

Ben Webb, Modeller Caretaker