I used Modeller in the past to create small loops from scratch. Now I have a protein-DNA complex, and I have to build and model 4 loops of my protein.
My first problem is that MODELLER does not support DNA, so I have to strip the DNA from the PDB and use the protein alone.
Considering the problem that I have to "paste" the DNA to the modelled structure, I use the options
a.md_level= None to keep the rest of the protein fixed and the option a.loop.md_level to model the loops.
Now, these two commands work on three of the 4 loops, but I have another problem now: I have a very long loop to build (35 aa) and using these two commands, MODELLER create a very strange hairpin loop to connect these two positions. I know which will be the correct structure of this loop because I have a similar protein that present this loop, and the modelling is wrong.
Finally, if I don't usage a.md_level=None, MODELLER builds a very good loop, but all the protein is in a complete different position with respect of the starting one.