gelly Documentation : Introduction
Copyright © 2006, 2007, 2008 Global Phasing Ltd.
All rights reserved.
- gelly is the name of the refinement module included as part of the BUSTER
- It is also supplied with BUSTER/autoBUSTER as a separate executable that can
be used to refine with the GEOMETRY function only (no X-ray term).
This is useful for fixing geometry restraint violations caused by manual rebuilding.
The regularisation will cause bond lengths, angles,
short contacts to be adjusted to the "ideal" values. It is easiest to use this separate
executable by using the gelly_refine wrapper as this will
do all the restraint file setup, coordinate conversion, weight assignment in an identical manner
- gelly is based on the geometry program supplied with TNT, written by
Lynn Ten Eyck and Dale
Tronrud. It replaces the geometry, shift and ncs programs
in BUSTER/autoBUSTER operation.
- gelly enables a much more powerful optimisation method to be used for X-ray crystallographic
- BUSTER/autoBUSTER using gelly (BUSTER-GELLY-TNT) will perform very much better
than the old BUSTER. The R-factor will drop much faster and to lower final values.
Feedback from users is very positive.
- If you are refining too many degrees of freedom then you will also notice that
the gelly optimiser quickly widens the Rfactor Rfree gap. This is not a problem
with the optimiser: instead it indicates that you are trying adjust more parameters
than your data/current model justify. We are currently working hard on procedures
to help for this case.
- gelly takes a novel approach for rigid body parameters. This makes it possible
to mix Cartesian and rigid body parameters during a refinement optimisation.
We are currently exploring the potential of this method to help during
refinement / model completion.
- The gelly optimisation module can normally be used without worrying about what is
- Often autoBUSTER will be using new features that gelly enables
in an automated fashion.
- However, you may wish to get access to features that have not yet
been scripted or to do things in a way distinct from what a script
- This manual provides the full details of the additional control cards
that gelly introduces. If you find anything confusing or need help in
other way please let us know.
- Gelly has been written primarily by Oliver S. Smart, with many contributions
from Tom Womack, Claus Flensburg, Wlodek Paciorek, Clemens Vonrhein, Maria
Brandl and Gerard Bricogne.
- It is under active development.
- We welcome comments and user feedback both good and bad.
- Please send feedback to firstname.lastname@example.org
The gelly refinement module is automatically run with the autoBUSTER
refinement has been requested. The autoBUSTER command line options
-Gelly and/or -RB
can be used to provide files containing user specified .Gelly cards.
Gelly cards all start with NOTE BUSTER and are used primarily to control
restraint setup and parameter selection. gelly also has some command line options.
The recommended way to run standalone gelly is by using the wrapper script
gelly_refine supplied with autoBUSTER. Separate
gelly_refine documentation is available.
To see the main options for this type
gelly_refine takes PDB files as an input, and automatically prepares
the .seq, .cor, .tnt files needed to run gelly. As
an example, for a simple run from the structure new_test.pdb
This will run a geometry-only optimisation of the structure and write results to
the directory test_dir. A geometry report for the initial and final
structures will be produced together with a summary of the optimisation.
gelly_refine -p new_test.pdb -d test_dir -o test_dir/gelly_refine_out.pdb
If you want to control the geometry function and/or parameter selection in gelly_refine use the
-l <filename> option. This can be used to
introduce a file of Gelly cards. For example
if you want to geometry regularise only part of a molecule that has been rebuilt before full refinement.
then produce a file with appropriate NOTE BUSTER_CONSTANT XYZ ...
cards and use this with the -l option.
- gelly command line options are used primarily to control detailed features of a run.
The controls can include things like the convergence gradient limit and how frequently pdb files are written.
- autoBUSTER now directly passes to gelly relevant command line options. (In the past the
autoBUSTER parameter BusterExtraArgs was used - this will still work).
- For the gelly_refine geometry regularization script. Command line not recognised by the wrapper script will be
passed directly to gelly.
For example, a useful combination of options is
gelly_refine -type one -screen_sigma 3.0 -p foo.pdb -d screenit -o /dev/null
which writes a summary of all the restraint violations worse than 3 sigma for foo.pdb
- The most useful gelly command line options are are
||automatically detect NCS and impose LSSR restraints for these.
||use weight w for -autoncs LSSR restraints.
|| impose target LSSR restraints to file.pdb.
||use weight w for -target LSSR restraints.
||Swap equivalent atoms to improve NCS and/or target restraints
||Distance cutoff d for special position handling
||Set the gradient convergence limit to terminate a small cycle. Terminate the optimisation when the
RMS gradient drops below n units|
||Stop the optimisation when the RMSd movement of the molecule is more than n Å|
||Write out a pdb file every n small cycles|
Set the level of reporting produced by gelly. The default level is -verbose 2;
verbose level 4 produces enormous lists, for example every inter-atom bond and the full
list of atoms in each COMBINE and CONSTANT set, but is useful
for bug reports.
||Adjust screen output so that it report the worst NS violations of each kind of restraint (default 100)|
||Report up to NS restraint violations worse than s sigmas (default 3.0)|
||gelly_refine only - Just calculate the initial function values and display a report, rather than optimize|
||gelly_refine only - Allow up to n optimisation steps (within autoBUSTER use -nsmall instead). |
Page Author: Oliver S. Smart
Please send feedback to: email@example.com
Last modification: 16.02.11