On 04/25/2012 12:16 PM, Jérôme Hénin wrote: > I am trying to do loop refinement on a multi-chain protein. My problem > is that Modeller tries to anchor C-ter loops to the N-ter of the next > chain, yielding unrealistic starting configurations, and making the > whole refinement fail. > > How can I instruct Modeller that the C-ter and the following N-ter > "loops" are to be modeled separately?
Modeling of a loop at the terminus is unlikely to give satisfactory results, since you really only have one anchor. The unanchored end is just going to flop around - most likely what you'll end up with is a disordered region stretching away from the protein.
But if I understand your question, you are trying to model a loop at the C terminus of one chain at the same time as a loop at the N terminus of the next chain. Modeller's default starting conformation for loop modeling places the loop atoms on a line between the nearest residues that are not being modeled (in sequence) - check out the *.IL*.pdb file. As you've observed, this will join the termini if you try to model two loops at the termini of neighboring chains. Without a defined terminus, this is Modeller's "best guess". If you can do better (e.g. the original PDB structure), override the build_ini_loop method: http://salilab.org/modeller/9.10/manual/node90.html
Ben Webb, Modeller Caretaker