You are here

Loop modeling: KIC with fragments producing unusual models

2 posts / 0 new
Last post
Loop modeling: KIC with fragments producing unusual models

Dear Friends,

I have used KIC with fragment protocol for loop modeling as follows
-s test.pdb
-loops:loop_file test.loops
-loops:frag_files frags/test.200.9mers frags/test.200.3mers none
-loops:frag_sizes 9 3 1
-loops:remodel perturb_kic_with_fragments
-loops:refine refine_kic_with_fragments
-out:nstruct 50
-out:file:silent test.out/test.silent
-out:file:silent_struct_type binary
-mute protocols.looprelax.FragmentPerturber
-mute core.fragments.ConstantLengthFragSet
-mute protocols.loop_modeling.loggers.ProgressBar

I am using rosetta_2014wk35_bundle.
The loop file without and with extension designated as follows
1. LOOP 147 161 160 0 0
2. LOOP 147 161 160 0 1
My output have been low energy structures with unusual secondary structure and very long bond lengths for the loop modeled region (pdb001_without extension) with a loop file without extension. While with loop extension the secondary structure looks normal but atomic detail has C-O distance of ~66 Å for each residue in the loop model region see pdb002_with_extension.
Could someone help me to figure out what is causing this. I have used the CCD and Next generation KIC protocol ( with -loops:frag_sizes and -loops:frag_files) call removed) and I do get properly modeled loops see pdb003_KIC

PS: Example pdb files below need to be renamed with a .gz extension and then uncompressed as they have been compressed using gzip

Thank for your time


pdb001_without_extension.pdb139.96 KB
pdb002_with_extension.pdb139.97 KB
pdb003_KIC.pdb139.92 KB
Post Situation: 
Mon, 2015-06-01 12:06

Looking at the code, it looks like at one point there is a specific check for extended loops, and it deliberately skips rebuilding loops that are not extended. I'm guessing in the models that look strange you're running up against this, resulting in loops which are not fully remodeled.

The fix is simply to skip doing the non-extended runs for the protocols which result in badly modeled loops under those conditions. (Fragment insertion should be thorough, anyway, so there wouldn't be an appreciable difference between doing with/without loop extension anyway.)

Fri, 2015-06-19 14:11