Hello,
I'm using Rosetta 3.4 and I want to make a homology model of a GPCR protein -> I know very well how the model should look like.
This is the input I'm using:
/opt/soft/rosetta3.4/rosetta_source/bin/minirosetta.linuxgccrelease \
-database /opt/soft/rosetta3.4/rosetta_database \
-in:file:fasta mc4.fasta \
-in:file:fullatom \
-in:file:native s1p.pdb \
-in:file:template_pdb s1p.pdb \
-in:file:psipred_ss2 mc4.psipred_ss2 \
-frag3 mc4_03 \
-frag9 mc4_09 \
-out:file:fullatom \
-loops:frag_files mc4_09 mc4_03 none \
-loops:frag_sizes 9 3 1 \
-in:file:fullatom \
-idealize_after_loop_close \
-loops:extended \
-loops:build_initial \
-loops:remodel quick_ccd \
-loops:refine refine_ccd \
-in:file:alignment alignment.aln \
-cm:aln_format general \
-out:nstruct 1 \
-out:file:silent mc4.silent.out \
-out:output \
-overwrite \
-ignore_zero_occupancy false
The template pdb is s1p.txt and the model is mc4_S_001.txt.
What has gone wrong? Any ideas?
| Attachment | Size |
|---|---|
 input.txt | 732 bytes |
| s1p.txt | 167.06 KB |
| mc4_S_001.txt | 146.5 KB |
"The" model? One of the common mistakes is not understanding that the abinito/homology modeling programs (well, all Rosetta programs, but abinitio/homology modeling in particular) are stochastic protocols. You can't just run them for one output structure and expect the first output structure to be good. Typically you should make thousands of structures, and then do clustering and energy filtering to pick out the best structure. ( See http://www.meilerlab.org/index.php/jobs/resources -> Rosetta Tutorials for a introduction/overview of the process.)
The other issue you're going to run into is the fact that it looks like you're modeling a trans-membrane helical bundle. By default Rosetta (like nature) relies on hydrophobic packing to stabilize structures. With membrane proteins, you don't have the apolar inside/polar outside to help organize things, so Rosetta is a bit behind the curve. The constraints to the template structure may correct for this, but if increasing the number of output structures doesn't help, you might want to look into incorporating the membrane into the scoring. See http://www.rosettacommons.org/manuals/archive/rosetta3.4_user_guide/db/d... for the description of membrane abinito.
That said, there's some concern that the membrane scoring for Rosetta isn't all that great. (For example, there was a membrane target in CASP10, and modeling it with membrane scoring didn't give anything good, but modeling it with the as-yet-unpublished Baker lab CASP10 protocol did give good results.) I'd recommend using the regular homology modeling protocol to start with (but with a larger number of structures), and only try the membrane protocol adjustments if that isn't working well.