This document describes comparative modeling using the minirosetta applocation. For a newer approach for performing comparative modeling in Rosetta, see RosettaCM
Author: James Thompson
This document was last updated on November 21st, 2010 by James Thompson (tex@uw.edu) . The PI is David Baker (dabaker@uw.edu) . The Rosetta comparative modeling protocol was developed primarily by the following members of the Rosetta Commons:
The code for running comparative modeling is in rosetta/main/source/src/protocols/comparative_modeling and rosetta/main/source/src/protocols/loops . See the rosetta/tests/integration/tests/threading directory for an example command-line and input files.
An introductory tutorial on comparative modeling can be found here.
This application was developed to build structural models of proteins using one or more known structures as templates for modeling.
The Rosetta comparative modeling approach is divided into four steps. The first step is generation of alignments to one or more template structures. We currently use external programs for this purpose, including BLAST, PSI-BLAST, HMMER and HHSearch. For more information on the alignment problem as it relates to comparative modeling problem, see Chivian and Baker (2006). The second step is generation of an incomplete model based on the template structure by copying coordinates over the aligned regions and rebuilding the missing parts using loop modeling. For more information, see the documentation for the loop modeling application. The third step is a full-atom refinement of the protein model using the Rosetta full-atom energy function. The final step is a selection of the models using clustering. File formats and command-line flags are detailed below in the Options section below).
The current standard protocol is to generate 10,000 separate models using the protocol detailed in this document, select the lowest 10% of models by energy, and then choose clusters using the "Cluster" application."
Fasta file. Contains the amino acid protein sequence in fasta format.
Fragment files. Generate structural fragment libraries using either the publicly available webserver ( http://robetta.bakerlab.org/fragmentsubmit.jsp ) or a local installation of the rosetta_fragments package.
Native structure (optional). The native PDB structure may be used for benchmarking. When used, the RMSD to native is calculated for each model and provided as an extra column in the score line.
Psipred secondary structure prediction psipred_ss2 file (optional). The Psipred secondary structure prediction file can be used to pick chainbreak points for loop-modeling, and is optional.
A template pdb file for threading
A sequence alignment file
Comparative modeling can be run using the minirosetta application with the following flags (to list all relevant commands, run with -help option):
./bin/minirosetta.linuxgccrelease
-run:protocol threading
-in:file:fasta t288_.fasta Query FASTA sequence
-in:file:fullatom
-out:file:fullatom
-loops:frag_files aat288_09_05.200_v1_3.gz aat288_03_05.200_v1_3.gz none Fragment files
-loops:frag_sizes 9 3 1 Number of residues in each fragment file
-in:file:native native.pdb Native structure (optional)
-in:file:psipred_ss2 t288_.psipred_ss2 PSIPRED-SS2 File (optional)
-in:file:fullatom
-idealize_after_loop_close Idealize structure after closing loops
-loops:extended
-loops:build_initial
-loops:remodel quick_ccd Method for rebuilding loops
-loops:relax fastrelax Protocol for full-atom refinement
-relax:fastrelax_repeats 8 Number of full-atom refinement cycles
-in:file:alignment t288_.result.filt.valid Alignment file
-cm:aln_format grishin Alignment file format
#-cm:aln_format general if general format is used
-in:file:template_pdb 1be9A.pdb List of template PDBs
-database path/to/rosetta/main/database Path to rosetta database
-nstruct 1 Number of output structures
-out:file:silent t288_.silent.out Use silent file output, use filename after this flag, default=default.out
(or -out:pdb) Use PDB file output, default=false
-out:path /my/path Path where PDB output files will be written to, default '.'The "grishin" file format for sequence alignments is listed below:
## t288_ 1be9A_4
# hhsearch_3 33
scores_from_program: 0.000000 0.998400
2 VPGKVTLQKDAQNLIGISIGGGAQYCPCLYIVQVFDNTPAALDGTVAAGDEITGVNGRSIKGKTKVEVAKMIQEVKGEVTIHYNKLQ
9 EPRRIVIHRGS-TGLGFNIIGGED-GEGIFISFILAGGPADLSGELRKGDQILSVNGVDLRNASHEQAAIALKNAGQTVTIIAQYKP
--The first two lines represent the identifier of the query and template sequences, and 1be9A.pdb must provided on the command-line with the -in:file:template_pdb option (listed above). The string "1be9A_4" should be a unique identifier for the sequence alignment. Alignment identifiers are stored in silent-files, and each structure in a silent-file should store which alignment was used as a template for model building.
If you provide alignment file in general format use the flag -cm:aln_format general The general format looks like:
score 123.456 t288_ 1 VIAFRCPRSFMDQPLAEHFSPFRVQHMDLSNS------VIEVSTL 1be9A 1 ILSLRRSLSYVIQGMANIESLNLSGCYNLTDNGLGHAFVQEIGSL
1be9A.pdb must provided on the command-line
See the AbinitioRelax extract options and AbinitioRelax cluster options for information on how to extract PDBs and cluster silent-files from comparative modeling.