autoBUSTER Example Application: basic NCS setup


  • This example shows how the command line options can be used to easily exploit NCS information in refinement. 1osg the structure of the complex between BAFF with the peptide bhpBR3 is used as an example.
    • The structure was solved at 3.0Å resolution by Starovasnik and coworkers pubmed entry
    • details of 1osg pdb at ebi
    • The structure is composed of two protein BAFF trimers related by an NCS 2-fold axis. Each of the protein subunits binds a cyclic bhpBR3 peptide.
    • The pdb structure is well built and has been sensibly refined with refmac including using conventional restraints on NCS (RMSD-based medium weighted). The pdb entry quotes a R value of 0.213 and a Free R 0.286.
    • It provides a nice example with interesting NCS.

Starting Files

  • The structure factors as supplied by the pdb: 1osg.mtz.
  • Initial pdb file for these runs: 1osg_stripHOH_Brefine.pdb.
    • This was produced from the 1osg pdb entry.The water and Mg ions were stripped out as they are questionable. Then to be able to demonstrate the effect of positional refinement the B factors were refined with atoms fixed to their pdb positions. All temperature factors set to 80, then B-factor refinement was performed keeping atomic positions fixed (job 1osgBasicNCS_000_control_pdb_refine_justB_2009). Note that individual B-factors but under tight restraints were used, as this is sensible at this resolution (see BusterFAQ)

Initial runs, control and -autoncs_noprune

  • As a control autobuster refinement is applied without any NCS restraints. The run can be done with the shell script 1osgBasicNCS_001_control_noncs_2009. (explanation of shell scripts used here).
  • To apply LSSR positional NCS restraints the job 1osgBasicNCS_002_autoncs_initial_2009 is used. This uses the autobuster command line options:
    • -autoncs_noprune automated NCS procedure is to be used. This will identify NCS and use the new LSSR restraints on position. No restraints are applied to temperature factors. Because the option is invoked with -autoncs_noprune rather than -autoncs the "prune out" procedure will not be used at this stage. LSSR restraints plateau and so allow differences to occur. However, at the final stages of refinement LSSR restraints can disrupt regions that are distinct but close. The automated "prune out" procedure can help with these see below.
    • -sim_swap_equiv_plus this will try to swapping equivalent atoms in residues such as PHE and ASP to improve the NCS agreement between the copies. The -sim_swap_equiv_plus procedure extends this allowing the flipping of GLN, ASN and HIS side chains and so could disrupt H-bond networks. Use at your own risk and check the results!

Initial Results

  • Details of the NCS recognized in the 002 run can be found in the LIST.html files produced in the run. The file for the first big cycle can be examined by:
firefox ab_runs/1osgBasicNCS_002_autoncs_initial_2009_result/01-BUSTER/Cycle-1/LIST.html
  • The procedure identifies NCS for the protein chains A to F as well as the peptide G to L. The -sim_swap_equiv_plus procedure swaps 65 residues in the first cycle.
  • Comparing the initial runs in terms of refinement statistics and geometry quality measures (calculated with MolProbity
structure Rwork/Rfree MolProbity Ramachandran favored MolProbity score
pdb# 0.1845/0.2427 94.52% 2.81
001_control_noncs 0.1690/0.2469 95.39% 2.59
002_autoncs_initial 0.1839/0.2243 96.49% 2.22
    • pdb#: 1osg pdb entry after B-factor refinement with autobuster (atom positions kept fixed).
    • Rfree/Rwork progress graph:
    • 1osgBasicNCS_001_002_cf2009.png
  • So without any ncs the Rfree slightly increases from the pdb position (that had weak RMS-based restraints). The Rfree-Rwork gap widens. The Molprobity Score improves now the 2009 improved geometry function is used.
  • But with -autoncs_noprune Rfree is dropped by 1.8% and the Molprobity scores are markedly improved.

Follow on run - with pruning

  • The 002 -autoncs_noprune run did not allow any pruning of the LSSR restraints. After this it is sensible to allow the procedure to turn "prune out" regions that are clearly distinct or where LSSR restaints are pulling the most.
  • This can be done by the job 1osgBasicNCS_003_autoncs_followon_2009 that starts from the pdb output of the 002 run. In this job:
    • option -autoncs is used.
    • but option -autoncs_noprune is not. Thus turning on pruning.
    • option -sim_swap_equiv_plus is also not used. Once this has been used once it is advisable to this off as the swaps will already have been done.

Follow on results

structure Rwork/Rfree MolProbity Ramachandran favored MolProbity score
pdb# 0.1845/0.2427 94.52% 2.81
001_control_noncs 0.1690/0.2469 95.39% 2.59
002_autoncs_initial 0.1839/0.2243 96.49% 2.22
003_followon 0.1808/0.2234 96.38% 2.23
    • Rfree/Rwork progress graph:
    • 1osgBasicNCS_001_002_003_cf2007.png
  • It can be seen that the use of the -autoncs option allows an improvement in the Rfree over the unrestrained by over 2% with a much better Rfree-Rwork gap. The geometry measures are also better.
  • Details of which residues are pruned can be found in the LIST.html files produced in the run (see the log file 1osgBasicNCS_003_autoncs_followon_2007.log for locations).
  • The pruning results in a slight drop in Rfree compared to the 002. But the most important benefit is often found in the maps.
    • In this case pruning now has little effect on the map difference.
    • But in results with a higher autoncs weight in the difference density around residue VAL A|227 was found in an -autoncs_noprune run: 1osgBasicNCS_002_autoncs_initial_coot_A227.png
    • This was eliminated in a comparable run -autoncs (with pruning) as this and surrounding residues are pruned out from the LSSR restraints: 1osgBasicNCS_003_autoncs_followon_coot_A227.png

Model improvement from the results of these runs.

  • What does the 003 run reveal?
    1. Running visualise-geometry-coot ab_runs/1osgBasicNCS_004_TLSfrom_003_result provides a quick way to scroll through geometry issues. It is not immediately obvious how to fix C|222 ASP but the side chain of D|155 THR should be rotated to hydrogen bond to D|152 ASP (like in the other 5 copies of the protein). Otherwise the list is dominated by CYS residues....
    2. Looking at the difference map the strongest features are around the disulfide bond in the each of the protein (difference map contoured at six sigma): 1osgBasicNCS_003_autoncs_followon_coot_disulfideF.png. This is likely to be due to the partial reduction of the disulfide bonds due to radiation damage. The use of -autoncs strengthens difference map peaks. How to deal with the partial reduction of a disulfide is the subject of another wiki example application.
    3. I think the refined BUSTER maps reveal an extra copy of the cyclic beta hair pin peptide bound at a crystal contact. Further refinement with TLS and -autoncs brings out the density for the extra peptide, as discussed in another example.

Result pdb and mtz files for each run

  • For the impatient
job pdb file mtz file
1osgBasicNCS_000_control_pdb_refine_justB_2009 1osgBasicNCS_000_control_pdb_refine_justB_2009_refine.pdb 1osgBasicNCS_000_control_pdb_refine_justB_2009_refine.mtz
1osgBasicNCS_001_control_noncs_2009 1osgBasicNCS_001_control_noncs_2009_refine.pdb 1osgBasicNCS_001_control_noncs_2009_refine.mtz
1osgBasicNCS_002_autoncs_initial_2009 1osgBasicNCS_002_autoncs_initial_2009_refine.pdb 1osgBasicNCS_002_autoncs_initial_2009_refine.mtz
1osgBasicNCS_003_autoncs_followon_2009 1osgBasicNCS_003_autoncs_followon_2009_refine.pdb 1osgBasicNCS_003_autoncs_followon_2009_refine.mtz

Page by Oliver Smart original version 2 May 2008. Revised for 2009 release with improved geometry function 20 July 2009. Address problems, corrections and clarifications to