[phenixbb] merohedral twinned refinement

Maia Cherney chern at ualberta.ca
Mon Jan 18 07:08:31 PST 2010


This looks similar to my problem crystal. It turned out that the crystal 
had a statistical disorder. Randy Read helped me with it. In the end, 
the crystal was abandoned.

Below is what Randy found out about my crystal

"The difficult crystal is difficult because there is some kind of 
statistical disorder. if I solve the structure using either 3HYW or the 
other P6222 structure you just sent, I get two similar, but 
significantly different, solutions. So I tried tricking Phaser into 
finding 2 half-weight copies, by telling it to search for 2 copies, 
saying that there are 2 copies of the sequence in the asymmetric unit, 
and turning off the packing check.  Then it finds both solutions 
simultaneously, with a much better LLG and R-factor than the single copy 
solutions.  (The R-factor from Phaser is 36% to 2.5A, without accounting 
properly for bulk solvent, when I use your other P6222 model.)

The fact that it's happier with both copies being present simultaneously 
tells me that this is disorder rather than twinning, because in the case 
of twinning the data are explained better by either single copy but not 
by both together.  (In terms of Pattersons, there are no vectors between 
the two copies if the structure is twinned, but there are if the 
structure is statistically disordered.)

 The centers of mass of the two copies differ by about 3A, but it's not 
a pure translation, because they also differ by about 5 degrees in 
orientation.

I tried refining the relative occupancies of the two copies in 
phenix.refine, but it looks like they are there in basically equal 
proportions, as the occupancies refine to 0.52 and 0.48.

If you didn't have the other crystals, then it might be worth trying to 
extract the maximum information from this structure possible, but it's 
probably not worth while.  I might play with it from a theoretical point 
of view, to see what it can teach us for other difficult structures 
(where people might not be lucky enough to have another crystal form). "




JXQI wrote:
> Dr.Zwart,
>  
> Thanks for the reply.
>  
> 2. The systematic absences and the Rsym support the P3121.I already 
> mentioned the model fit very with the map if I use the P3121.
>  
> 3. If I try P3,the Matthews coeffs tell 2 or 3 molecules in the 
> ASU,the molecules will crash.Besides,the overall FOM ( lower than 0.4) 
> and the R/Rfree values (about 0.45/0.56) suggest that the P3 is the 
> wrong sg.
>  
> 4. TLS refinement also did not imrove the results.
>  
> 5. There are totally 88 aa in my protein, I could build 86 aa clearly.
>  
> The intensity ststistics is attached in the message.
>  
> J.X. QI
>  
> ************************************************************************
>  
>      Shell            I/Sigma in resolution shells:
>   Lower Upper      % of of reflections with I / Sigma less than
>   limit limit     0     1     2     3     5    10    20   >20  total
>   50.00  4.95   0.9   1.5   1.9   2.4   2.8   4.0   9.5  90.5  100.0
>    4.95  3.93   0.0   0.2   0.3   0.5   0.8   2.0   6.6  93.4  100.0
>    3.93  3.44   0.2   0.3   0.6   1.0   2.5   5.4  15.4  84.6  100.0
>    3.44  3.12   0.0   0.2   0.5   0.7   2.5   7.6  22.7  77.3  100.0
>    3.12  2.90   0.3   1.0   2.1   3.2   7.2  17.6  41.4  58.6  100.0
>    2.90  2.73   0.0   0.7   2.0   3.2   9.9  20.8  52.9  47.1  100.0
>    2.73  2.59   0.0   1.4   2.9   5.7  12.4  30.0  60.8  39.2  100.0
>    2.59  2.48   0.7   1.6   4.7   7.7  16.3  36.0  71.5  28.5  100.0
>    2.48  2.38   0.6   2.1   5.0  10.2  21.9  42.0  76.2  23.8  100.0
>    2.38  2.30   0.8   2.3   7.4  12.7  25.3  49.4  83.4  16.6  100.0
>  All hkl        0.4   1.1   2.7   4.7  10.0  21.1  43.4  56.6  100.0
>  
>  
>  
>  
>  Summary of reflections intensities and R-factors by shells
>      R linear = SUM ( ABS(I - <I>)) / SUM (I)
>      R square = SUM ( (I - <I>) ** 2) / SUM (I ** 2)
>      Chi**2   = SUM ( (I - <I>) ** 2) / (Error ** 2 * N / (N-1) ) )
>      In all sums single measurements are excluded
>  
>  Shell Lower Upper Average      Average     Norm. Linear Square
>  limit    Angstrom       I   error   stat. Chi**2  R-fac  R-fac
>       50.00   4.95   619.8    13.8     3.5  1.039  0.046  0.050
>        4.95   3.93   834.6    17.5     5.0  1.156  0.048  0.052
>        3.93   3.44   521.9    14.6     4.5  1.026  0.059  0.062
>        3.44   3.12   333.7    11.1     4.1  1.073  0.071  0.068
>        3.12   2.90   172.5     7.1     3.6  1.153  0.098  0.096
>        2.90   2.73   131.4     6.0     3.4  1.256  0.116  0.108
>        2.73   2.59   102.8     5.2     3.4  1.306  0.140  0.130
>        2.59   2.48    77.7     4.6     3.4  1.381  0.173  0.158
>        2.48   2.38    64.1     4.3     3.3  1.355  0.195  0.177
>        2.38   2.30    54.7     4.1     3.3  1.345  0.221  0.197
>   All reflections    297.8     8.9     3.7  1.209  0.071  0.059
>  
>      Intensities of systematic absences
>       h   k   l  Intensity     Sigma   I/Sigma
>  
>       0   0   2      -0.2       0.4      -0.5
>       0   0   4       0.1       1.0       0.1
>       0   0   5       0.4       1.4       0.3
>       0   0   7       1.1       2.5       0.4
>       0   0   8       0.4       2.1       0.2
>       0   0  10       7.3       3.4       2.2
>       0   0  11       6.0       3.5       1.7
>       0   0  13      -2.3       4.9      -0.5
>       0   0  14       6.1       4.3       1.4
>       0   0  16       8.1       5.1       1.6
>       0   0  17       3.9       4.9       0.8
>       0   0  19       0.8       4.5       0.2
>       0   0  20       4.0       4.6       0.9
> *********************************************************************
>  
>  
> ** 
> ** 
>  
> 1. How do the intensity statistic look like?
> 2. How did you choose P3121?
> 3. Did you try solving and refining in P3?
> 4. TLS?
> 5. How complete is your model
>
>
>
>
>
> 2010/1/17 JXQI <jxqi at mail.im.ac.cn 
> <http://phenix-online.org/mailman/listinfo/phenixbb>>:
> >/ Dear all,
> />/
> />/ I have a 2.3A data set processed with HKL2000. The space group is 
> P3121 with
> />/ 63% of solvent content and 1 molecule in the ASU. I used a 40% 
> homology
> />/ model to do molecular replacement. Both the Phaser and Molrep gave 
> similar
> />/ solution with good packing. After further refinement with phenix, 
> the built
> />/ model fit well with the 2mFo-DFc map. However the R/Rfree were 
> stuck to
> />/ 0.34/0.38.
> />/
> />/ I rechedked the data set with xtriage and found the data set is 
> merohedral
> />/ twinned (with twin law"-h,-k,l", the estimated twin fraction is 
> about 0.05).
> />/ When I used the twin law in further refinement, the R/Rfree were only
> />/ decreased to 0.32/0.36. Any suggestions are appreciated, thanks!
> />/
> />/
> />/ J.X. QI
> />/ _______________________________________________
> />/ phenixbb mailing list
> />/ phenixbb at phenix-online.org 
> <http://phenix-online.org/mailman/listinfo/phenixbb>
> />/ http://phenix-online.org/mailman/listinfo/phenixbb
> /
> ------------------------------------------------------------------------
>
> _______________________________________________
> phenixbb mailing list
> phenixbb at phenix-online.org
> http://phenix-online.org/mailman/listinfo/phenixbb
>   



More information about the phenixbb mailing list