[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [ccp4bb]: Denzo v. Mosflm
*** For details on how to be removed from this list visit the ***
*** CCP4 home page http://www.ccp4.ac.uk ***
On Thu, 16 Jan 2003, James Stroud wrote:
> *** For details on how to be removed from this list visit the ***
> *** CCP4 home page http://www.ccp4.ac.uk ***
>
> I am refining a structure with synchrotron data that has been processed by
> denzo_3d/scalepack. I have refined it nearly as far as I can go to 3.2 A,
> but my stats look like I should be able to get more out of this data (at
> least push it to 3.0 Angstroms). The crystal is somewhat mosaic (0.67 deg),
> diffracts anisotropically, and has a fairly large P2(1) unit cell (101.034
> 94.764 112.878 Ang.). The I/sig and R-facs in my outer bin looks hopeful:
>
> Shell Lower Upper Average Average Norm. Linear Square
> limit Angstrom I error stat. Chi**2 R-fac R-fac
> 20.00 6.40 6165.8 188.1 66.7 1.160 0.038 0.050
> 6.40 5.11 3175.2 114.7 67.1 1.056 0.052 0.056
> 5.11 4.47 3479.6 139.1 88.4 1.187 0.057 0.087
> 4.47 4.06 3006.3 187.6 106.4 0.793 0.070 0.065
> 4.06 3.78 2029.3 153.7 107.1 1.038 0.105 0.101
> 3.78 3.55 1525.0 148.2 118.3 1.103 0.140 0.116
> 3.55 3.38 1241.7 159.4 137.2 1.234 0.160 0.121
> 3.38 3.23 1119.2 175.1 155.8 1.151 0.136 0.087
> 3.23 3.11 898.8 180.7 166.4 1.147 0.144 0.097
> 3.11 3.00 429.6 163.0 158.9 1.183 0.314 0.253
> All reflections 2503.2 159.9 112.0 1.086 0.075 0.069
>
>
> But my completeness in the outer bins is disasterous:
>
> 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
> 20.00 6.40 0.6 1.1 1.8 2.5 3.8 7.5 19.1 74.8 93.9
> 6.40 5.11 0.9 2.4 3.6 5.2 8.3 16.0 34.2 60.8 95.0
> 5.11 4.47 1.5 3.1 4.7 6.3 9.9 18.4 38.8 56.8 95.6
> 4.47 4.06 1.7 3.5 5.9 8.5 12.9 28.2 65.9 30.2 96.1
> 4.06 3.78 2.5 5.9 9.8 14.4 23.1 43.7 77.1 19.3 96.4
> 3.78 3.55 1.8 6.0 12.3 19.5 32.6 58.0 85.8 10.8 96.5
> 3.55 3.38 3.5 9.4 18.9 29.7 45.7 69.7 86.9 7.4 94.2
> 3.38 3.23 5.4 15.6 27.9 37.7 52.0 66.4 76.4 5.9 82.3
> 3.23 3.11 7.1 15.8 25.8 34.4 44.0 52.5 57.8 4.4 62.2
> 3.11 3.00 6.2 14.6 22.6 27.9 33.6 38.8 41.7 0.9 42.6
> All hkl 3.1 7.7 13.3 18.5 26.5 39.8 58.2 27.3 85.5
>
> It seems like scalepack is tossing reflections. Is there a reason why? I
> picked these resolution limits by inspecting as denzo integrated. There did
> not appear to be a great number of reflections tossed during integration.
> This appears to be happening during scaling, so I don't believe it is
> overlap. Since I used denzo_3d, the fact that I have many partials (1.0
> degree oscillations) should not hurt, should it? Eg.:
>
> type good intns weak overlap backgr others total outside tail
> ovfl prob
> partials 461 0 2142 8 3 28 2642 1 0
> whole 48 0 187 0 1 0 236 0
> total 509 0 2329 8 4 28 2878 1
>
> Scalepack seems to do a fine job with the lower resolution bins where there
> are only partials also. I am not sure where the reflections went because the
> reject file only has 429 measurements tossed (anomalous was not used). So my
> questions are
>
> 1. what is denzo/scalepack doing here?
> 2. will mosflm (which I have yet to learn) handle this situation better?
>
> I will appreciate any help.
>
> James
>
James,
I've had problems with the unit cell dimensions meandering during
refinement. If the change is significant enough, scalepack can start
mis-indexing the hi-res reflections, and subsequently tosses them out.
Check to see if the unit cell dimensions are wandering during refinement
in the HKL2000 gui. If this is the case, you might try taking the refined
unit cell dimensions from scalepack, puting those into denzo and then not
fitting the unit cell dimensions on the next pass of denzo. Best of luck
Scott Reid
tsr@duke.edu