I would like to use MODELLER to optimize some models that I have generated as part of a loop prediction protocol. The loop conformations were built onto crystal structures downloaded onto the PDB; ideally I only want the loop region to be optimized. I thought the code given on the following web page would be what I should use, however it doesn't seem to do what I want:

This still generates the initial 'straight' conformation that the MODELLER loop modelling algorithm starts from; it doesn't start from the loop conformation in the file I provided.

Is anyone able to shed some light on this for me?