DM for ncs averaging (CCP4: General)

NAME

dm_ncs_averaging - dm for ncs averaging

AVERAGING REQUIREMENTS

In order to perform an averaging calculation you need the following information:
  1. Averaging Matrix.
    The averaging, or 'RT' operators are a rotation and translation which describe the mapping between one area of density and another. For DM they can be given as matrices in CCP4 or O format, or as Euler/Polar angles and translations. They are always given in orthogonal angstrom coordinates.
  2. Averaging Mask.
    This covers the region of the unit cell that obeys the non-crystallographic symmetry. Note that it does not cover a unit cell or an asymmetric unit, it does not obey the cell repeat, and it does not obey the crystallographic symmetry. It is fundamentally different from a solvent mask. If you do not understand this you had better go back to the textbooks, or hope that the 'automask' facility works right first time.

MAKING AVERAGING MATRICES

The most likely place to find your averaging operators is from running a self-rotation and translation function, typically in AMORE. This will give the operator you need in Euler angles, which may be used in DM.

You must get the matrices the right way round. For example, if you calculate a matrix which maps an A chain onto a B chain, then the averaging mask must cover the volume occupied by the A chain.

Alternatively you can make your averaging matrices if you can build some corresponding fragments of structure in symmetry related molecules. Heavy atom coordinates can also be used in the determination of averaging matrices, especially if the rotation function has been solved. In this case you will need to run a program such as LSQKAB.

MAKING AVERAGING MATRICES USING LSQKAB

You can use the program LSQKAB from the CCP4 library to calculate the matrices relating one set of atoms to another. See the documentation of that program to find out more about writing the command script. Here is an example script which will find the matrix to map 125 residues from chain B onto chain A. You could also match e.g. chain A in one pdb file to another chain A in a different file. 

lsqkab                   
     REFRCD chmi.pdb \  
     WORKCD chmi.pdb \ 
     LSQOP  junk.pdb \
  << 'END-lsqkab'
OUTPUT  XYZ
FIT  WRESIDU MAIN 2 TO 126 WCHAIN B
MATCH RRESIDU 2 TO 126 RCHAIN A
END
'END-lsqkab'
The output of this run included the following information: 

          SUM DISPLACEMENTS**2 =  78.070
          SQRT(SUM DISPLACEMENTS**2)=   0.395
          AVERAGE DISPLACEMENT =   0.236
          MAXIMUM DISPLACEMENT =   5.001


      ROTATION MATRIX:
           -0.43671  0.05443  0.89796
           -0.62780  0.69647 -0.34754
           -0.64432 -0.71551 -0.26998
  TRANSLATION VECTOR IN AS   43.63491  38.05914  62.72586

 .........

  CROWTHER ALPHA BETA GAMMA  158.84190-105.66331 132.00317
  SPHERICAL POLARS OMEGA PHI CHI  113.28130 103.41944 120.33858
  DIRECTION COSINES OF ROTATION AXIS   -0.21318   0.89350  -0.39524
The small average displacement is a good indication that a correct match has been found. The symmetry operators may then be input to DM, either as matrices using the ROTA MATRIX and TRANS cards, or as angles and translations. Alternatively, the matrix may be transposed and input in O/RAVE format.

Thus in this case, the 'AVER' card in the DM command file was as follows (note that a second run of LSQKAB was used to determine the third symmetry, between chains A and C): 


AVER REF
ROTA POLAR  0.0  0.0  0.0
TRANS  0.0  0.0  0.0
AVER REF
ROTA MATRIX -0.43671  0.05443  0.89796     -0.62780  0.69647 -0.34754  -
            -0.64432 -0.71551 -0.26998
TRANS  43.635 38.059 62.726
AVER REF
ROTA MATRIX -0.42948 -0.62559 -0.65130      0.06496  0.69793 -0.71322  -
             0.90074 -0.34862 -0.25911
TRANS  82.989 15.401 -8.928
or 

AVER REF
ROTA POLAR  0.0  0.0  0.0
TRANS  0.0  0.0  0.0
AVER REF
ROTA EULER 158.84190 -105.66331 132.00317
TRANS  43.635 38.059 62.726
AVER REF
ROTA EULER  47.59828 -105.01736  21.15850
TRANS  82.989 15.401 -8.928
or 

AVER REF
ROTA POLAR  0.0  0.0  0.0
TRANS  0.0  0.0  0.0
AVER REF
ROTA POLAR  113.28130 103.41944 120.33858
TRANS  43.635 38.059 62.726
AVER REF
ROTA POLAR   66.58067 -76.78019 119.69176
TRANS  82.989 15.401 -8.928
or 

AVER REF
OMAT
  1.000  0.000  0.000
  0.000  1.000  0.000
  0.000  0.000  1.000
  0.000  0.000  0.000
AVER REF
OMAT
 -0.437 -0.628 -0.644
  0.054  0.696 -0.716
  0.898 -0.348 -0.270
 43.635 38.059 62.726
AVER REF
OMAT
 -0.429  0.065  0.901
 -0.626  0.698 -0.349
 -0.651 -0.713 -0.259
 82.989 15.401 -8.928

Making Averaging Masks

The easiest way to make an averaging mask is to let DM generate one by its built in local-correlation-masking code. However even this requires some thought. The program must pick a correlation cutoff to isolate the correct masked volume, so it needs to know how much of the map will agree with a rotated copy of itself. This is achieved with the 'NCSMASK NMER <nmer>' card.

You need to specify how many monomers will map onto matching density under the non-crystallographic symmetry transformations. In the simplest case - purely improper ncs - the answer is 1. In the case of purely proper ncs (e.g. 3-fold rotation, or 2-2-2 ncs), then it is the number of monomers in the crystallographic asymmetric unit (3 and 4 in these cases).

More complex cases occur when ncs operators are related to crystallographic operators, usually to build up larger multimers. For example, in Insulin there is a 2-fold ncs axis perpendicular to the crystallographic 3-fold and intersecting it, with the result that the entire hexamer obeys the noncrystallographic symmetry, so <nmer>=6. If fact this case is further complicated because neighbouring cells up and down the c-axis are related onto each other, so the CLIM card must also be used.

Given the complexity of the task, it is often better to make your own mask. The local-correlation map can also be generated by MAPROT and used in conjunction with graphics programs, MAPROT, and NCSMASK, to generate an averaging mask.

Making a mask from coordinates

The other common method for generating an averaging mask is from coordinates or bones. Coordinates are available when the model has been partly built, or when there is a molecular replacement solution.

Bones are generally generated and then pruned in 'O'. The output bones file can be turned into a pseudo-pdb file using the utility BONDES2PDB.

In either case, the .pdb file can be converted directly into an averaging mask using the NCSMASK program.

AUTHOR

Dr. Kevin Cowtan, Protein Structure Group, University of York, England