[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[ccp4bb]: Re: [o-info] refinement problem



***  For details on how to be removed from this list visit the  ***
***    CCP4 home page http://www.dl.ac.uk/CCP/CCP4/main.html    ***

"Yoder, Marilyn" wrote:
> 
> Dear All,
> 
> I'm having a refinement problem and would appreciate any suggestions.
> 
> I have a well-refined, well-behaved 2.2 A MAD data set that the structure was
> solved with Se atoms to a crystallographic R of  21.2% and a free R of  25.2%
> (269 of a possible 271 amino acids, one phospholipid, 55 waters).  The crystals
> are P21, a = 43.914 , b = 73.773 , c = 48.185 , b = 114.732.
> 
> My problem is in refining a higher resolution native data set, to 1.9 A.  The
> data is fairly isomorphous with the MAD data set, P21, a = 43.718 , b = 73.184
> , c = 47.427 , b =  help analysis data quality.
.  All data has been processed with denzo/scalepack
> (including the MAD data).  For the high resolution refinement, I took the model
> from the MAD data, did a rigid body refinement, simulated annealing, and a
> B-individual refinement, manually rebuilt a little, ran a conjugate gradient
> minimization -- but my crystallographic R is stuck at about 27%, and the free R
> is about 31%.  The 2fomfc maps calculated after this refinement are beautiful.
> Truly, without exaggeration.  I can contour to about 2 sig and see holes in my
> aromatic residues, and water molecules are clearly visible.  Carbonyl oxygen
> bulges are readily apparent throughout the molecule.  I can automatically add
> about 100 water molecules, and my crystallographic R drops to 25.0% and a free R
> of 28.8%.
> 
> I have tried numerous refinement strategies, including refining with and without
> the MAD data set, cutting the resolution to 50-2.2A, cutting the resolution to
> 10-1.9A, and nothing seems to get me beyond this 25% crystallographic R.  The
> maps look, to me at least, way to good to have this high of an R.  The B's
> refined with the MAD data and the native data are very comparable.  OMIT maps
> showed no obvious problem with the model.
> 
> My gut feeling is there is a problem with the data, but it looks good to me.
> I'm including the end two tables from the final scalepack run.  The only thing I
> can think of with this data, is that it is a synchrotron data set, and during
> that data collection run there were considerable problems with the beam (it kept
> dumping prematurely).
> 
> I would appreciate any comments or suggestions.
> 
>      Shell            I/Sigma in resolution shells:
>   Lower Upper      % of reflections with I / Sigma less than
>   limit limit     0     1     2     3     5    10    20   >20  total
>   50.00  4.09   2.7   4.3   5.1   5.5   6.0   7.9  12.8  85.6   98.4
>    4.09  3.25   0.2   0.6   1.1   2.4   4.2   8.6  19.1  80.8   99.9
>    3.25  2.84   1.2   2.5   4.8   6.6  11.4  21.2  46.5  53.4  100.0
>    2.84  2.58   2.2   4.8   7.8  11.1  18.7  36.8 100.0   0.0  100.0
>    2.58  2.39   3.2   7.3  11.9  17.3  26.4  51.5  99.9   0.0   99.9
>    2.39  2.25   4.3  10.3  18.1  24.4  35.5  70.0  99.8   0.0   99.8
>    2.25  2.14   4.6  11.6  21.8  31.5  48.8  99.6  99.6   0.0   99.6
>    2.14  2.05   7.2  16.2  27.5  37.6  56.2  97.9  97.9   0.0   97.9
>    2.05  1.97   9.5  19.2  29.5  40.4  58.3  94.5  94.5   0.0   94.5
>    1.97  1.90   8.2  20.7  33.3  44.5  66.3  88.4  88.4   0.0   88.4
>  All hkl        4.3   9.7  16.0  22.0  33.0  57.4  75.6  22.3   97.9
> 
>                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   4.09  2614.5    72.3    36.3  0.981  0.028  0.033
>        4.09   3.25  2132.1    61.3    34.0  1.295  0.037  0.040
>        3.25   2.84   935.9    42.2    23.6  1.243  0.056  0.055
>        2.84   2.58   565.2    40.9    20.0  0.991  0.075  0.071
>        2.58   2.39   382.9    35.8    19.4  1.008  0.102  0.093
>        2.39   2.25   305.4    36.2    20.9  1.106  0.134  0.121
>        2.25   2.14   253.9    41.8    22.6  1.011  0.193  0.000
>        2.14   2.05   199.1    38.8    23.7  1.027  0.196  0.170
>        2.05   1.97   174.9    37.4    24.4  1.024  0.207  0.177
>        1.97   1.90   138.4    37.9    24.0  0.979  0.243  0.221
>   All reflections    787.7    44.7    24.9  1.072  0.062  0.044
> 
> Thanks,
> Marilyn Yoder
> 
> Marilyn D. Yoder                                               yoderm@umkc.edu
> Division of Cell Biology and Biophysics           phone: 816-235-1986
> School of Biological Sciences                          fax:      816-235-1503
> University of Missouri-Kansas City
> 5007 Rockhill Rd.
> Kansas City,  MO  64110-2499
> 
> _______________________________________________
> o-info mailing list
> o-info@o-info.imsb.au.dk
> http://o-info.imsb.au.dk/mailman/listinfo/o-info


 IT is hard to say without more information.
And harder still as I dont know CNS diagnostics..

 If you import scalepack data via the GUI you get some very useful plots
which help analyse data qualitty. Look at wilson plot ; cumulative
intensities, fall off of <F> and anisotropic falloff..

Again with REFMAC5 you get plots which might help diagnose scaling
problems etc.

 Do you have 2 molecules in the asymmetric unit? Pseudo translation NCS
can increase R values greatly


And did you transfer the FreeR set from the first data set to the
second?

 Maybe not essential but can help..
Eleanor