Is there any way to ignore some loops on threading? I'm trying to build a model based in a template which is about 30% identical to the query sequence. The sequence has lots of loops ranging from few aa to 80 aa. I was wondering if would be possible to ignore the biggest loop(s) at first and loopmodel them in a separate protocol.
Looks like the some loops are not closing so as the program finishes without build any model.
Perhaps someone could suggest a better approach.
What protocol are you using to remodel your loops? It should be as simple as simply omitting those regions you don't want to remodel from your loops file. ("Loops" in loop remodeling is a bit of a misnomer - it doesn't necessarily correspond to secondary structure assignment. Instead, a "loop" is simply a region of the backbone you wish to remodel - regardless of it's predicted secondary structure. So you can omit non-alpha, non-beta regions from your loops file if you want. You can also have alpha or beta regions in your loops file, if that is what you want to remodel.)
My first try was the basic homology modeling by threading forcing extended loops, independent of loop input file. I used -random_grow_loops_by flags. This way I was planning remodel loops by using the loopmodel protocol. It didn't work. Any model was built and minirosetta.mpi stoped without any reasonable explanation (-out:level 400). Secondly, I tried to input the loop file which also didn't help.
I'm afraid I'm still not quite understanding which protocol you're trying to use. Could you post the exact commandline(s) of the command that you ran which isn't doing what you want? It's best if you could also attach the files needed for that run. (particularly options files, but also loops files, input PDBs, etc. - you may need to rename them to end with .txt in order to post them on the forums. Fragment files are too big to post to the forum, but I'm not sure I'd need to see them quite yet.)
Here goes the flag file I used in my first try. As you see, there isn't noting new to default. miniroseta.mpi doesn't finish the work and stops without complain anything.
In the second try a used a loop file omitting the biggest loops. I also tried other things but without success.
EDIT: I've downgraded to my old rosetta install (rosetta_2014wk05) and things are running non-stopping for one hour (asked -nstruct 200), but without build any model so far. My idea is to make things run and decide what to do afterwards. Perhaps someone may suggest a better approach. Don't know what is best: build model ignoring loops in a first run (threading protocol) and remodel loops on a second step (using loopmodel) or build models and loops at the same protocol.
The minirosetta threading protocol automatically chooses loops based on your alignment, and there really isn't any way to turn off particular loops. (Other protocols use the loops file to specify which loops to model, but the minirosetta protocol does it automatically and doesn't take a loop file.)
One possibility is simply just to edit your modeled sequence to remove the inserted residues you don't want to model. No insertion - no gap - no remodeling. You can then run the protocol on the loops you do want to remodel, and then later on run on the other residues. The drawback of this is that for each input sequence edit you make you need to redo all your input files, including your fragment files, such that they correspond to the sequence you're modeling at that stage.
Alternatively, you could change your modeling protocol to one of the two stage ones, where you thread the sequence first, and then do loop remodeling afterwards. (See the various "Homology or Loop Modelling", "Comparative Modeling, Loop Building and Clustering" etc. materials at Rosetta Tutorials at http://meilerlab.org/index.php/jobs/resources for an example of the two-stage protocol in use.) In those protocols you specify which loops to remodel with a loop file, and as such can remodel only a selected loop, while leaving the rest of the loops in an unmodeled state.
Doing the multi-stage approach will allow you more control over how things are modeled.