NCS PROBS

axel t. brunger brunger at LAPLACE.CSB.YALE.EDU
Mon Jun 12 22:49:56 EST 1995


A few comments:

1. The two top grid points of the translation function have very
close "T" values.  Are two points related by symmetry?

2. The NCS operators will be applied when using PC-refinement.

3.  You can use NCS symmetry in rigid body refinement.  However,
the NCS relationships will be fixed.  In other words, only movements
will be allowed that conserve the NCS matrices and translations.
Refinement of these relationships is currently not possible.  A grid
search might be an alternative.

4. Gerard Klejwegt is a good person to ask about the conversion
from RAVE to X-PLOR matrices.

5. We have a version of mapman for very large maps.  Again, Gerard
is the person to ask.

Axel Brunger

>
> Dear ALL CRYSTALLOGRAPHERS,
> 	I have recently entered into the world of crystallography  and I am
> working on the structure of an icosahedral virus. There are a number of
> problems I have been encountering and I was hoping that some of you have
> had experience in dealing with these situations.
>      My questions regard both the use of strict NCS in structure determination,
> as well as the use of RAVE for electron density averaging.  I would appreciate
> any help you could offer!
>   1)  The virus I am working with is in the P21212 space group and 30 of its 60
> protomers are in the asymmetric unit, each of the 30 are related to each other
> by NCS.
> The virus sits on the crystallographic 2-fold axis with 30 of its protomers
> related to the other 30 by a two-fold rotation about z.
>      So, there are only two degrees of freedom, rotation about, and translation
> along the z axis. Based on similar structures, the virion  rotation should be
> about 3 degrees (or less) about the z axis, without being zero degrees that
> would put the virus in a body centered space group.
>      The translation along the z axis should be around 0.25 or 0.75.
> Using XPLOR, when I do cross-rotation searches, the top peak is usually
> sphere=(90 90 0) and variations thereof, so my assumption is that the
> rotation is very slight and is masked by the crystallographic two-fold about z.
> When I go directly into PC refinement without any rotation, with a CA model of
> the asymmetric unit without using strict NCS, the mininum is
> 	 Corr<E(obs)^2, E(calc)^2> = 0.097  and R=0.5863
>   with  group= 1 rot/tran=(    0.00    0.00   -2.86    0.14    0.00    0.00)
> and Etotal =9027.025   grad(E)=0.064      E(NCS )=0.000      E(XREF)=9027.025
>       And after a search of x=y=0 and z=-0.05 to  0.76, the top peaks are:
> T=(  0.000   0.000 284.006) TF=(   0.000   0.000   0.749) T= 0.1113 P= 0.1304
> T=(  0.000   0.000  95.445) TF=(   0.000   0.000   0.252) T= 0.1098 P= 0.1303
> T=(  0.000   0.000  91.207) TF=(   0.000   0.000   0.240) T= 0.0796 P= 0.1303
>
>       I did not use NCS strict in the search because I was not sure if the
> NCS operators are updated when the search is carried out. If they are not
> updated, then the asymmetric unit is not described correctly. I then put the
> output coords from the translation search into rigid body refinement and the
> Rwork went down to 0.4737 and Rfree=0.4763.
>
>        The first problem is that when I compare the pdb coords that come out of
> the translation search they are related to the input coords by a 2.86 rotation
> about x, not z (determined using lsq_ex in O). BUT, the minimum's position from
> PC refinement is euler=(0 0 -2.86)=2.86 rotation about -z not x.
> Why do you think this is occurring and is it correct to assume that the coords
> are positioned  according to the rot/tran statement in the refinement summary?
>
>       Secondly, the rigid body refinement was not limited to variations in
> rotation about z, that should be limited due to the NCS (i.e., there must
> be rotation and translation only about z).
> Is there any way to limit the refinement to two-degrees of freedom?
>
>       Do you think that I am not seeing a rotation solution because of the
> possibility that the crystallographic two-fold is masking the NCS 2-fold
> rotation?  If I continue to use the coords from the translation solution,
> can I use strict NCS in rigid body refinement (with a skew matrix to shift
> the NCS operators to the position of the starting model) and safely assume that
> the molecule is described correctly even after small rotations and translations
> after each cycle of refinement (i.e., do the NCS operators get updated? )
>
>         2)  Once the correct orientation is found I would like to begin
> averaging (perhaps even before to help identifiy the correct orientation
> and position).  I noticed that there is a RAVE edition for viruses compiled
> on a DEC/ALPHA.  Is there such a version for SGIs or can I use 4d_big
> versions?
> Also, what do you think is the best way to refine NCS operators?
>
> 	3)  When I calculate the original NCS operators using lsq_ex in O,
> and write-out the .lsq_rt matrices to a file, am I correct in assuming that
> the format written to the file can be used directly in XPLOR?
>
> Example of  the O output is:
> 2       .LSQ_RT_A1ONA2            R         12 (8(x,f8.5))
>   0.30901  0.80902 -0.50000 -0.80902  0.50000  0.30902  0.50000  0.30902
>   0.80902  0.00018 -0.00009 -0.00001
>
> and then the ncs strict statement should be:
>
>   xncsrel  {===>  NCS # 2  }
>       matrix=(0.30901  0.80902 -0.50000)
>              (-0.80902  0.50000  0.30902)
>              (0.50000  0.30902  0.80902)
>       translation=(0. 0. 0.)
>     end
> However, when I use CELLO to convert the O format to XPLOR the matrix is
> actually:
>    xncsrel
>     matrix      = (      0.309010    -0.809020     0.500000 )
>                   (      0.809020     0.500000     0.309020 )
>                   (     -0.500000     0.309020     0.809020 )
>     translation = (        0.0002      -0.0001       0.0000 )
>    end
> Do I have the conventions confused??
>
>     4)   The last question is unrelated to this specific series of problems.
> When I calculate a map in XPLOR using a cubic cell of a=b=c=445.1, there seems
> to be a problem in MAPMAN that does not allow such large unit cell dimensions.
> O gives an error saying that the extent of the map is outside the box limits.
> When I decrease the cell dimensions in MAPMAN, I am able to display a map
> but it doesn't seem to correlate well with my structure.
> Is this problem overcome in a different version of mapman or another type of
> conversion program?
>
> 	If you're reading this portion of this letter that means that you've
> probably gotten through it.  I thank you for even reading the whole thing.
> Any of your insights regarding averaging, NCS, and structure determination are
> welcome.
>
> Sincerely,
> Dawn Resnick
> Center for Advanced Biotechnology & Medicine
> Rutgers University
> Piscataway, NJ 08854
>
> resnick at mbcl.rutgers.edu
> 	
>
>



More information about the X-plor mailing list