Queen of Spades on Biowulf

Qs (Queen of Spades)

Queen of Spades is a "brute force" style molecular replacement program which uses a method based on a reverse Monte Carlo minimisation of the conventional crystallographic R-factor in the 6n-dimensional space defined by the rotational and translational parameters of the n molecules. Because all parameters of all molecules are determined simultaneously, this algorithm should improve the signal-to-noise ratio in difficult cases involving high crystallographic/non-crystallographic symmetry in tightly packed crystal forms.

Because a successful run of this application can last weeks or months on a single processor, it is best used in parallel fashion on the biowulf cluster. It should not be the first application to try for molecular replacement (try AMoRe instead, for example), but rather as a last resort.

Version: 1.3

Author: Nicholas M. Glykos

How to use

Minimally, you need three files for a single model per asymmetric unit: a pdb file containing the coordinates of the model, the data in hkl format, and a parameter input file.

Qs_test.pdb:

ATOM      1  N   LYS     1     -11.460  31.361  22.771  1.00 20.48           N  
ATOM      2  CA  LYS     1     -11.083  30.241  21.810  1.00 20.62           C  
ATOM      3  C   LYS     1     -11.021  29.061  22.764  1.00 20.27           C  
ATOM      4  O   LYS     1     -10.316  29.368  23.686  1.00 20.31           O  
ATOM      5  CB  LYS     1      -9.766  30.428  21.114  1.00 20.34           C  
...

Qs_test.hkl:

  -34   0   1      173.60      1.00
  -34   0   2      548.05      1.00
  -34   0   3      137.69      1.00
  -34   0   4      229.47      1.00
  -34   2   1      188.31      1.00
...

Qs_test.in:

#
# Example script for testing Qs
#
TARGET          R-FACTOR
CYCLES          1
STEPS           100000
STARTING_T      0.01500
FINAL_T         0.00500
INFO            1000
NOISE_ADDED     0.20
RESOLUTION      150.0 4.0
AMPLIT_CUTOFF   500.0
SIGMA_CUTOFF    0.0
RANDOM_SELECT   1.0
FREE            0.20
MODEL           example.pdb
DATA            example.hkl
GLOBAL_B        20.0
MOLECULES       1
SEED            357539
SCALECELL       4.0
MAXGRIDSPACING  1.0
SCMODE          wilson
INTERPOLATION   linear
POSTSCRIPT      colour
CELL            103.900   38.700   34.000   90.000  100.600   90.000
GROUP           5

Once all these files are created and are in a single directory, the Qs job can be launched onto the cluster.

Here is a sample qsub script (named Qs.qsub):

#!/bin/bash
#PBS -N Qs_test
#PBS -e Qs_test.err
#PBS -o Qs_test.log
cd $PBS_O_WORKDIR
/usr/local/bin/Qs Qs_test.in > Qs_test.log

The Qs.qsub script can be submitted to the cluster with the command

$ qsub -l nodes=1 Qs.qsub

Parallel use

Typically, a search is done with multiple cycles of minimization. These cycles can be broken into seperate jobs and run on individual processors to save time and utilize the power of the cluster. Create a series of subdirectories corresponding to each run (for example mod_1, mod_2, mod_3, etc.), copying the files needed to run into each. Set the CYCLES parameter to 1. Set the SEED parameter is different for each input file, otherwise the minimzations will all be identical. Then either submit each with single qsub scripts, or with the swarm command.

Here is an example swarm file (named Qs.swarm):

cd /data/user/mod_1; /usr/local/bin/Qs test_1.in > test_1.log
cd /data/user/mod_2; /usr/local/bin/Qs test_2.in > test_2.log
cd /data/user/mod_3; /usr/local/bin/Qs test_3.in > test_3.log
cd /data/user/mod_4; /usr/local/bin/Qs test_4.in > test_4.log
cd /data/user/mod_5; /usr/local/bin/Qs test_5.in > test_5.log
cd /data/user/mod_6; /usr/local/bin/Qs test_6.in > test_6.log
cd /data/user/mod_7; /usr/local/bin/Qs test_7.in > test_7.log
cd /data/user/mod_8; /usr/local/bin/Qs test_8.in > test_8.log

The Qs.swarm script can be submitted to the cluster with the command

$ swarm -l nodes=1 -f Qs.swarm

Output

Aside from the redirected output file, a series of PDB files containing the final coordinates of your search model(s) from each minimisation is generated.

The program will also create a PDB file containing the coordinates of all symmetry equivalent copies of all search models, which can be viewed directly with RasMol (to check for bad contacts). You will need an X11 connection to run RasMol. Type rasmol QsXXX_packing.pdb and then type set unitcell on followed by set stereo on, colour chain, and backbone. You should now have a coloured side-by-side stereo image of the contents of 2x2x2 unit cells corresponding to the XXXth minimisation.

Additionally, a series of postscript files are generated. The files are colour postscript files containing projections of the model unit cell and central sections through the molecular transform(s), and are completely useless from the structure determination point of view (see keyword POSTSCRIPT for more information). Viewing postscript files - requires an X11 connection, gs myfile.ps. Type man gs for more information.

More information and finer details

Qs is pretty straightforward to use, but like ice hockey and bubble gum, it also takes time to master the finer details. For those who want more information see the Qs Online Help.

Warnings

Qs is ultimately dependent on physical memory to work well. Large unit cells and high resolution data require a great deal of memory. If your job fails to progress (as seen by either no output, or output stopped at some point), the node may be caught in a continuous swap cycle. Memory usage of a node can be seen in the live biowulf monitor. This may result in the node hanging, and your job will not complete. If this is the case, stop the job and resubmit with a lower resolution range.