These macros are to be included in the February 2011 release of BUSTER
In response to a number of requests, we have written and been testing some macro procedures to carry out short autoBUSTER runs.
The aim of these protocols is to allow a quick refinement to be carried out after manual building or tweaking of a model (eg in Coot) or ligand addition. In other words to quickly validate relatively minor changes to a model.
These protocols are still under development but we are reasonably happy with the results thus far and have made a few of these available to you now for your own use. We would appreciate your feedback on how they perform for you and any ot her suggestions you have.
The two macros provided (they can be listed using the command refine -M list) are ShortRun and ShortRunVoid. They can be used with
refine -M ShortRun or refine -M ShortRunVoid.
ShortRun, will carry out a single big cycle with a maximum of 20 small cycles. ShortRunVoid, our preferred procedure, is similar but will also apply a void correction and recalculate the solvent mask.
The scope and use of these protocols is limited - they are designed for use during the process of model completion/improvement, NOT for initial refinements on incomplete or only partially built structures, after major model reconstruction or initial addition of waters (after which we would STRONGLY recommend a full autoBUSTER run) - and as such, a few prerequisites and limitations should be borne in mind when using them.
The major factors likely to affect the outcome of a short refinement are the geometry restraints. BUSTER uses different geometry restraints to other refinement packages such as phenix and refmac. Therefore, before using these short protocols we would STRONGLY recommend that a full BUSTER refinement should have been performed in order to correct the geometry. A full refinement will also have the advantage of calculating an "appropriate" x-ray weight which will be used in subsequent runs. Although the overall weight and geometry of modified/added residues will change on modification of the model, they are unlikely to have deviated enough to compromise a limited refinement.
Depending on the scope of any changes made to the model following a full refinement, the short refinement procedure may well not reach "convergence". However, it should generate improved geometry of the rebuilt section and an improved map for validation.
In order to optimise the speed, a number of checks that autoBUSTER normally carries out, such as mtz and gelly sanity checking have been disabled. Having run a full autoBUSTER refinement prior to using these protocols, these checks will have carried out and verified previously thus there should be no problem with the input files.
It is also important to note that use of either of these short run macros is NOT compatible with the concurrent use water addition or ANY of the other supplied macros.
If TLS refinement has previously been carried out and TLS definitions are present in the pdb header then these can be applied (but not refined) by specifying the -TLS option. When using either of the short run macros, we would recommend that the only parameters that should be specified (other than the input pdb and mtz files) are -autoncs, -l commands if external dictionaries are required and -TLS if previously refined TLS parameters are to be applied. No other macros should be specified.
Amongst the test cases that have been used in testing these protocols is the sulfotransferase SULT1A3 used for the autoBUSTER tutorial - see AutoBusterExampleIntro wiki page.
Steps from E through to K in this tutorial were carried out. The start point was a molecular replacement model with only ~50% sequence identity and with some missing regions as well as two unmodelled ligands. The sequence was corrected, badly modelled regions corrected, missing residues manually built and the ligands fitted in coot. In between each building step, the model was refined using the ShortRunVoid macro to validate each building step and to improve the map to allow correction of the next feature - 9 cycles of model building and short refinement in all.
The initial R/Rfree of the model was 36.3/36.7. This had dropped to 22.4/25.3 at the end of the final cycle. After a final full run of autoBUSTER (5 big cycles each with up to 100 small cycles) the final R/Rfree ended up at 19.1/23.4.
The maps at each stage were interpretable and each map was improved from the previous cycle (as were the R/Rfree and geometry).
In terms of time, autoBUSTER jobs with either of these two macros are very comparable to typical, single refinement jobs using refmac.
Page by Andrew Sharff original version 24 January 2011. Address problems, corrections and clarifications to buster-develop@globalphasing.com