Rosetta 3.2 Release Manual

Documentation comparative modeling of protein structures

James Thompson


This document was last updated on November 21st, 2010 by James Thompson <>. The PI is David Baker <>. The Rosetta comparative modeling protocol was developed primarily by the following members of the Rosetta Commons:

Code and Demo

The code for running comparative modeling is in src/protocols/comparative_modeling and src/protocols/loops. See the test/integration/tests/threading directory for an example command-line and input files.



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."

Input Files


Comparative modeling can be run using the minirosetta application with the following flags (to list all relevant commands, run with -help option):

-run:protocol threading

-in:file:fasta t288_.fasta                                                    Query FASTA sequence

-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)

-idealize_after_loop_close                                                    Idealize structure after closing loops

-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
-in:file:template_pdb 1be9A.pdb                                               List of template PDBs
-database ../minirosetta_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

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.

See the AbinitioRelax extract options and AbinitioRelax cluster options for information on how to extract PDBs and cluster silent-files from comparative modeling.

Generated on Wed Feb 16 16:04:25 2011 for Rosetta Projects by  doxygen 1.5.9

© Copyright Rosetta Commons Member Institutions. For more information, see