gelly provides a number of approaches to best exploit similarity information with restraints during refinement.
Similarity can be between different chains within the structure being refined.
Such similarity arises because of Non Crystallographic Symmetry (NCS) and is conventionally exploited using
restraints on positions that are based on the superposition of the related chains (or parts of the chain).
The restraints are sometimes known as "soft-NCS" but will be called "superposition-based NCS" here.
BUSTER provided superposition-based NCS restraints from their first releases using the TNT ncs program.
The gelly module has its own superposition-based NCS restraint routines.
The gelly code uses RMSD superposition routines from:
Coutsias, E.A., Seok, C., Dill, K.A.(2004) "Using quaternions to calculate RMSD", J. Comput. Chem., 25:1849-1857.
The gelly routines essentially reproduced the TNT approach with some corrections (to derivatives) and improvements (more fine set selection).
The problem with superposition-based NCS restraints is that in practice their use tends to be involved because
it is necessary to adjust restraints to take account of NCS violations and domain type motions.
NCS violations are where the structures of the NCS partners are clearly distinct. This often arises because of different
crystal packing environments for part of the structure. If a conventional superposition-based NCS restraint is used then NCS violations
can only arise during refinement
if the atoms involved are removed from the restraint ("pruned out"). If distinct parts of the structure are modeled into density then
but not pruned out from superposition-based NCS restraints then the next round of refinement will be dominated by moving the
structures to agree with one another (not what is wanted!).
An alternative approach is to relax the restraints by setting a low weight or high sigma.
Although NCS violations are allowed by this approach much of the benefit for genuinely related parts of the structure is lost.
Domain type motions are also prevented by the naive application of superposition-based NCS restraints. It is necessary
to separate the domains into different NCS sets and provide a separate restraint groups.
Because of the complications in application there is a tendency for NCS information not to be used in refinement,
despite the very clear benefits of using NCS restraints in refinement.
Publications describing the method in full are in preparation.
This documentation provides information about how to use the
restraints in practice with this release of BUSTER (May 2008).
LSSR restraints reduce the need for manual intervention in using NCS restraints during refinement.
Because of this simple command line arguments for their activation have been introduced.
The automated options provide a simple approach that should satisfy many requirements. However, they may need some
manual adjustment.
Because of this it is also possible for users to precisely tailor the restraints in a detailed manner
using .Gelly type cards.
It should be noted that if users prefer to use conventional superposition-based NCS restraints these are still
available for both positions and temperature factors. NCS setups from previous releases will work in exactly the same manner as before.
NCS restraints allow the exploitation of the similarity between two or more chains within a structure.
But suppose instead of having NCS we know that the structure being refined is closely related to a
structure that has already been solved.
We call the known structure a "target structure". The target structure is provided by the user
as a pdb-format file. It remains fixed during the refinement.
Restraints between the refined structure (or part of it) to the fixed target structure are very
much the same as NCS restraints between two or more chains in the refined structure.
LSSR restraints can be used for target restraints.
A command line option facilitates the most common type of application.
It is also possible to use superposition-based restraints to a target structure. There is no need to superpose the
target structure onto the refined structure.
Page Author: Oliver S. Smart
Please send feedback to: buster-develop@globalphasing.com
Last modification: 10.07.08