Hi,
What would be the most suitable procedure to model ab initio a long linker (~60 aas) between two protein domains, which structures and relative positions are knowns? This linker is missing in an X-ray structure of the protein, and has no homology in the pdb.
Is it possible to model the whole protein by homology on the X-ray template, build the linker, and then optimize only the linker while "freezing" the rest of the protein?
Thanks in advance,
Isaure
Post Situation:
You have several options here, depending on your system.
The first question is, does the domain fold? If you expect the domain to fold stably, fold it by itself and try to assemble the quaternary structure later. If the domain does not have any stable structure, then that limits what information you can learn about it via Rosetta. Rosetta almost exclusively focuses on modeling stable, crystallizable structures (the scorefunction is parameterized from crystal structures); if your domain doesn't adopt a single folded conformation, then Rosetta will have trouble modeling it.
The second question is, should the two known domains of the system move with respect to each other (it appears you've said no)? If they move, you have a flexible region, which may be treated with FloppyTail. If you want them to stay static, you have a loop (or a folded domain insert), which you should treat with loop modeling, or maybe hacking up some code and doing domain insertion.
60 residues much, much too long for loop modeling - loop quality falls off noticeably after 12 residues.
I believe it to be possible to do the sort of modeling you envision, but it's not directly accessible via command line. You can use Relax with a custom MoveMap via RosettaScripts, but it won't keep your endpoints in place. There was a demo published from the 2011 RosettaCON collection for "partial ab initio" where some of the structure was known and some was not, but I don't think that was for a middle region between two proteins.