I am interested in mutating some loop residues together with the residues of the beta-strand , the loop is connected with. In such case what has to be done?.. Does fixedbb design will work in this case ??
Also , I have another query regarding Rosetta Scoring. I want to test all the 20 possibility of amino acids at two loop residues (20*20=400) of a beta hairpin . In that case will there be any large change in energies of a pair of combination of amino acids ??
In addition to this how can I create such a library of 400 sequences , testing the possibility of all 20 amino acids at 2 positions ??
Fixbb will work on pretty much any fixed backbone. Certainly any standard protein will work; Rosetta may fail on weird stuff like helix staples or cyclized proteins. A design region that is partly sheet and partly loop is no problem; the backbone stays fixed.
If you sample all 400 combinations of residues at two positions, there will certainly be at least a few with a large energy change. If the positions are exposed, then probably most combinations will have nearly the same energy, because the sidechains won't interact much. You may want to turn off the reference energy term for this experiment - since you are doing the design exhaustively it will make the results harder to interpret (normally it biases designs towards "realistic" sequences, but you want _all_ sequences).
I think the pmut_scan application is a good choice for this. (http://www.rosettacommons.org/manuals/archive/rosetta3.3_user_guide/d0/d...) - particularly the -mutant_list option. Failing that, use a script (write one in python or perl) to generate the 400 resfiles and run them sequentially - it won't take long (maybe two hours on one slow processor?)
I am not able to locate pmut_scan protocol anywhere at the mentioned location in the documentation (http://www.rosettacommons.org/manuals/archive/rosetta3.3_user_guide/d0/d...)
You're right...its documentation shouldn't be in the 3.3 manual. The bottom of the page says it's supposed to release in 3.4. I've been told a 3.4 beta will release in a few weeks, but you can also just create the resfiles with a script as I suggested.
I was able to generate the resfile using a perl script. Thanks for your advice. I have one doubt regarding the scoring in rosetta fixedbb design and KIC loop modelling. Actually I have modelled a two residue tight turn using fixed bb design for those 400 sequences that I generated. This actually produced sequences with large energy variation. So, I want to know is it right to model the turn residues, using fixed bb design or shall I go for loop modeling. Will loop modeling will also generate the same scoring trend as that was generated by fixed bb design ??
It would be appropriate to do limited loop modeling on the redesigned loops to account for the sequence change. Rosetta is likely to produce a similar scoring trend once the loops are remodeled, but not precisely the same order. (If it was exactly the same, why waste time doing it?) I think you can get away with a small amount of sampling here (since you have so many inputs to start out with) - if you can afford nstruct 1000 do it, but I think you'd get useful results out of nstruct 50 or even nstruct 10.
Actually I tried modeling those two loop residues with loop_modeling protocol, the protocol is not mutating the residues of the loop that are mentioned in resfile. I have also checked that the application is reading the resfile by incorporating a bad line in resfile, which resulted in an error. That's why I am asking whether is correct to model the loop residues with fixedbb design or not and what could be wrong with the loop modeling application that it is not mutating the residues ??
You should have gotten 400 output PDBs, with your 400 different sequences, as output from fixbb. Use those as inputs to loop modeling (no resfile necessary).
You shouldn't do loop modeling on only two residues - Kinematic modeling requires a minimum of three and I think it works best with 5 to 8. I guess CCD would work but the fragment requirements are onerous...
The loop size is of 4 residue (VNGH) and I am giving option for only two residue to mutate (NG) in KIC loop modeling protocol. For generating fragments , I am using Robetta sever. Is that fine ??
You shouldn't need fragments for KIC only loop modeling (it depends somewhat on the options you use). Other than that, 4 residues of loop is enough for KIC to run, and probably enough to get score discrimination of which sequences are worst and hopefully which are best.
But the problem here is that the KIC loop modeling protocol doesnot mutate the residue as per the resfile listing. They always return the same sequence as that of the original protein sequence. What could be the reason for this ?
I guess the documentation is wrong. The whole loop modeling executable is a mess - it works fine if you know what flags to use, but most of the stuff it does isn't documented or is documented wrong.
You already have starting PDBs for the 400 sequences you want to model. Use those as the inputs for loop modeling and it won't matter if it can design or not since the sequences are already correct. If you don't want to do all 400, then just do the ones that score best by fixbb (and by your biophysicists' intuition).
Sorry to ask this but after using the those models which already have the mutated sequence for loop modeling, what would that do . If I am not wrong it will try to put the best rotamer for the amino acid and would give an energy score. But I have already got the energy score for those mutated sequence from fixedbb design. After checking with the literature the best scoring models have right sequence preference. That means the modeling has been done correctly. Then what would loop modeling can add additional to this ??
Loop modeling would allow the backbone to change in response to the mutations. Check out the KIC papers to see what I mean. If you already have the design sequences your looking for through fixedBB design and the lowest energy matches the sequence tolerance of the backbone, then cool. FixedBB design worked, and the energy function correctly predicted the best sequences. This doesn't happen very often, but it depends on the specific modeling problem and the constraints on the structure.
Lets say you are designing on a surface loop. It's 6 amino acids long. You do your 20x20 mutation matrix for two positions. The backbone may or may not change to accommodate the given mutation. It may just change slightly to optimize the loop and find a close lower energy state. (You could sample this by using the FastRelax protocol or Backrub after design) Sometimes though, depending on the mutation and the position/environment of the loop, the loop may change structure significantly. This is where you would need a higher degree of conformational sampling to get the right structure (and subsequently energy) of the loop.
As you can see, it depends on your specific modeling problem, as well as the number of mutations you make (uncertainty of backbone structure increases exponentially with the number of mutations ), and perhaps what resolution you're looking for in your final structures....
Hi, I am trying to model a four residue loop, using loop modeling KIC protocol. The loop consists of 4 residues. while executing the command , I am getting the following error :-
loops: ERROR invalid loop 101 105 103: End more than nres(24)
ERROR: LoopRebuild::ERROR Loop definition out of boundary
Here's the command that I am using :- ./loopmodel.linuxgccrelease -database /usr/local/rosetta3.3_bundles/rosetta_database/ -loops:remodel perturb_kic -loops:refine refine_kic -resfile resfile -loops:input_pdb s.pdb -in:file:fullatom -loops:loop_file loop_file.loop -out:file:fullatom -nstruct 100 -overwrite -out:prefix s
The pdb file contains only the beta-hairpin , numbered from 92-115. Here's the loop_file and resfile that I am using :-
LOOP 101 105 103 0 0
101 A PIKAA ALLAAxc
102 A PIKAA ALLAAxc
103 A PIKAA ALLAAxc
104 A PIKAA ALLAAxc
But when I use the full protein instead of a hairpin, the program is executing perfectly fine.
Please can anybody tell me where am I going wrong ??
You are trying to use PDB numbering in your loops file. The loops file takes Rosetta numbering (residues numbered from 1). If the first residue is 92, and there are no gaps, then you need to use loop 9 through 14, I think.
Thanks for your help. Actually I tried loop modeling for a 4 residue loop , giving chance for all 20 aa to occur (except Cys) at all the four positions. But every time rosetta is modeling Leu at all the four positions. I used loop_kic protocol for loop modeling ?? I thought of writing a script generating resfile for all possible combination of amino-acids for those 4 positions. But I am unable to write that. So is there any other way to accomplish this ??
I didn't look at it carefully before (since your previous post was about the loop file not working, but the stated resfile:
101 A PIKAA ALLAAxc
102 A PIKAA ALLAAxc
103 A PIKAA ALLAAxc
104 A PIKAA ALLAAxc
Is not going to work the way you want. PIKAA means "pick these amino acids" and is followed by the residues you want. ALLAAxc means "pick any of the 19 standard amino acids (excepting cysteine)". They aren't meant to be used together. I suspect it's interpreting the As and Ls in ALLAA to mean, "PIKAA with only A and L". Just remove the PIKAA from your resfile and try again. I don't know if it will work or not; the KIC documentation says it will...
I have a query regarding fixed bb design. Actually I have designed some 250 mutants for a beta-hairpin turn keeping . The turn was of two residues and I kept the backbone fixed using fixed bb design. After that I have calculated the dihedral angles for those two turn residues, I found that every mutant has the same set of dihedral angle. Why is it so ??
I'm not sure I understand the question - you did fixed backbone design (meaning the dihedrals, angles, and lengths don't change), and the dihedrals didn't change - what's the problem? If you want loop sampling, you need to run loop modeling after fixbb, as we've discussed...