refinement against anisotropic data

[Jiang Jiansheng]

> 1) I would really like to put an overall anisotropic B-factor on the
> ATOM side of the equation, but the only software I have found that will
> do this is an old version of PROLSQ.  Does anybody out there know of a
> better way?  I have tried correcting Fobs in XPLOR, with and without
> sigma-based weighting schemes, but my R-factor is hanging up much
> higher than it should.  


> 2) After anisotropic scaling of Fo to Fc in XPLOR, there are still
> systematic differences between the two.  Short of local scaling Fo to
> Fc, (EEK!) has anybody found a better solution?

Hi Phoebe,

We have already implemented this feature in the new version of X-PLOR.
We will be glad that some users give it a pre-trial with the new version.

   { anisotropic B scaling to alter the old scaling method  }
   { the linear scale constant is in "$K2" and the B tensor }
   { is in "$B2_11, $B2_22, $B2_33, $B2_12, $B2_13, $B2_23" }   

   multiscale
      anisotropic=true
      set1=fobs  k1=-1 b1=0
      set2=fcalc
      selection=(2.0 <= d <= 6.0 and 0.1 <= ampl(fobs))
   end

   { decalre a reciprocal object ANISOB to store the scaling factors }
   declare name=anisob domain=reciprocal type=real end
   do (anisob=$K2*exp(-( $B2_11*h*h*$astar*$astar
                              +$B2_22*k*k*$bstar*$bstar
                              +$B2_33*l*l*$cstar*$cstar
                           +2.0*$B2_12*h*k*$astar*$bstar
                           +2.0*$B2_13*h*l*$astar*$cstar
                           +2.0*$B2_23*k*l*$bstar*$cstar
                                 )/4.0)) ( all )
   { apply the anisotropic B scale to fcalc }
   do (fcalc=anisob*fcalc) ( all )

For details, please check with the x-plor www home page:
"http://xplor.csb.yale.edu", look for "group/people/jiansheng_jiang",
and look at the example files "anisoB_scale.*" in the subdirectory.
(the files may be copied onto the www pages by an overnight)

If the R value was increased by the anisotropic scaling, there
were probably something wrong.

Jiansheng Jiang.