I am trying to do ab initio prediction of a 62 aa protein. The models generated without any restraints do not seem to converge towards a particular structure (which is understandable). This protein has an indication of being structurally similar to another protein (36 aa long) for which I have already done the modeling and the model seems good. Since the sequence identity between them is very poor (~12%). I am skeptical in using this model for homology modeling but I would like to use this model to get structures similar to the already generated model.
So the question is that - Is it possible to use structure model (not in database) while fragment generation along with the usual database and use this fragment library in ab initio prediction?