Using FiltRest3D from the command line
To rank a set of decoys you can use a web interface or
install the program on your own workstation.
The models to be scored can be given in
Protein Databank file format,
Rosetta silent file format or
CABS' trajectory format.
Restraint file syntax
General restraint file syntax is:
DEFAULT_WEIGHT <- 1.0
There are following restraint types (restraint type name in parentheses):
- distance: dist;
- secondary structure assignment (helix,
flexible) and proportions (ss_composition).
- solvent accessibility: (access);
- protein shape: (edm).
If we want to have C-alpha at least 4.5 Angstroms apart and C-beta atoms within at most 7.3 Angstroms distance:
(V10)-(D34) (CA>4.5 CB<=7.3)
To restraint SG
atom of the residue C140 to be within 4 angstroms from
atom of residue E33:
C140_near_E33: (C140)-(E33) (SG-#OE1#<=4)
Secondary structure assignment and percentage composition
Secondary structure can be assigned to segments using helix, strand
and coil declarations.
coil ( c1: 201-205 )
Secondary structure percentage in the overall structure composition can
be described in the ss_composition block:
20 < helix < 50
15 < strand < 40
Above code declares between 20% to 50% of aminoacids in helices, and
between 15% to 40% of aminoacids in beta strands.
The example below declares unnamed aminoacid 45th as having between 40%
to 90% accessibility, Glycine 56th as having 30% to 60% accessibility,
and alanine 20th as having between 80% to 90% accessibility.
A20 80-90 weight<-1.5
Protein shape fit
'Situs' or '3SOM' strings
select the name of electron density matching program, and the last string
defines file name. Appropriate surface shape will be converted into electron
density first, using pdblur program from Situs.
[Installation process is described here.]
surface1: 'Situs' 'saxs_data.pdb'
surface2: '3SOM' 'remote_model_with_filler_atoms.pdb'
Conjunction and alternative
To declare composite restraint having score of minimum or maximum of two
other restraints, the following restraint blocks can be used:
Default restraint weight is 1.0, but it can be changed:
DEFAULT_WEIGHT <- 2
(V10)-(34-36) (CA<4.5 weight<-2 // C-alpha penalty will be multiplied by 2.0
CB<=7.3 weight<-1.5) // C-beta penalty will be multiplied by 1.5
Restraint weight selection
Careful choice of restraint weights is the most important part after
restraint selection. The suggested procedure is:
Design summary contribution for desired features (e.g. 40%
for core topology, 40% for active center geometry and 20% for binding
Give a 2-4x higher weight for low resolution feature components
(e.g. distance between whole groups of aminoacids) that are more likely
to be correct, and lesser weight for higher resolution components that
are less likely to be correct (e.g. atom-atom distances or single residue
exposure to solvent).
If you find a set of nearly good structures, check if some restraints
correlate better with your visual assessment - suboptimal weights may have
been chosen on the first try;
If you find no good structures, but very many models you may want
to use skyline.py search to find small set of
all structures which could be preferred for some weights. If no structure
is given by the skyline, you need to add different restraints and/or build
The downside of skyline.py is that the found
weights should be manually assessed for their plausibility and the structures
can be less reliably picked. The upside is that you may get the definitive
answer that the set of restraints is inadequate independently of choosen
weights, or you may see what features of the protein are harder to model.
The rules of a thumb for restraint weight selection is:
|restraint type||restraint type token||weight multiplier|
|for fuzzy distance restraints between long segments||dist||4.0-8.0|
|lots of alternatives||or||4.0|
|secondary structure segments||strand|
|for aminoacid-aminoacid or atom-atom distance restraints||dist||1.0|
|maximum penalty of many restraints||and||0.5|
|others||edm||not enough data collected|