have you tried the backrub public app?
this should be well suited to generate small localised ensembles of the backbone and repack the surrounding sidechains.
see recent papers by colin smith and tanja kortemme for reference.
PhD student, Baker group
> Looking for some guidance on taking a solved PDB structure and explore the different ensembles based on relaxing the backbone and working through possible backbone/rotamers adjustments and the corresponding pdb structure/energy score.
> The result should generate a collection of PDB structures that can then be used to find differences in the surface accessibility of the protein based on possible stable/flexible tertiary structures.