I've had some success using the Zhang server (I-TASSER) to generate a model for MR. The benefit ended up being finding a domain, which we hadn't known based on our amino acid sequence. Exciting times indeed.
This seems to be in the same vein, kayhi :
http://www.slideshare.net/ijstokes/wide-search-molecular-rep...
Will be used for my next project!
I guess, this story highlights quite elegantly the "protein folding problem" to be a canonical "P vs NP" problem! Exciting times!!