------------ CCP4 Newsletter - January 1997 ------------
Structural Biology (including Biocomputing) Programme, EMBL,
Meyerhofstrasse 1, D69012 Heidelberg
The standard method of using an image plate scanner is to record diffraction patterns over rotation ranges greatly in excess of the reflection rocking curve width. In this way one maximizes the density of information on a single image and reduces data collection and processing overheads. The disadvantage of this procedure is clear, namely that the signal to noise ratio must be worse (extra background is accumulating under the reflections when they are no longer in the reflecting condition). There is nothing to prevent one collecting mini-rotation frames on image plate scanners and perhaps one should consider more carefully whether or not to do this, especially as instrument duty cycles become shorter. That the use of mini-rotation frames with single-photon counters is advantageous is well established experimentally (W.Kabsch, Proceedings of the CCP4 Study Weekend on Data Collection and Processing, 1993, 63-70). For a MAR image plate, if one takes 0.25° frames and sums them to give 1° frames the data is better when the narrower frames are used for processing (P.A.Tucker, Joint CCP4 and ESF-EACMB Newsletter on Protein Crystallography. Number 28 (May 1993), 74-76). This experiment ignored the fact that for any analog detector (like an image plate) there must be some readout noise per image which, because this is summed as well, gives a poorer signal to noise ratio for the wider images than is realistic. To do a better experiment is simple, using a small ( 0.25 x 0.25 x 0.1 mm) crystal of tetragonal hen egg white lysozyme, three equivalent data sets were collected on a MAR image plate scanner using CuKa radiation. All data sets were collected over the same 45° under identical conditions except for the frame width and time. The first and last sets had frame widths of 0.2° measured over 40s. The second set had a frame width of 1.5° measured for 300s. All data sets were processed with XDS using parameterisation for the beam crossfire and crystal mosaicity empirically determined from the first data set. The results are summarized in the Table below and show clearly that narrow frame widths yield better data. Note that the smaller number of processed reflections for the second data set results from excluding reflections that were partially overlapped.
| Set 1 | Set 2 | Set 3 | |
|---|---|---|---|
| Frame width | 0.2° | 1.5° | 0.2° |
| Integrated reflections | 22841 | 18075 | 22847 |
| Outliers rejected | 225 | 1025 | 279 |
| Unique reflections | 6370 | 6048 | 6371 |
| Rsym as a function of resolution | |||
| 15.0-6.32 | 3.3 | 2.9 | 2.9 |
| 6.32-4.47 | 2.9 | 3.4 | 3.0 |
| 4.47-3.65 | 2.8 | 3.1 | 2.8 |
| 3.65-3.16 | 3.0 | 3.9 | 2.9 |
| 3.16-2.83 | 3.3 | 4.4 | 3.2 |
| 2.83-2.58 | 3.8 | 5.3 | 3.9 |
| 2.58-2.39 | 4.7 | 6.6 | 4.7 |
| 2.39-2.24 | 4.9 | 7.1 | 5.0 |
| 2.24-2.10 | 6.0 | 9.5 | 6.0 |
| % of reflections with I/s(I) > 12 | |||
| 15.0-6.32 | 91.3 | 89.6 | 91.6 |
| 6.32-4.47 | 94.3 | 91.3 | 94.4 |
| 4.47-3.65 | 95.3 | 92.6 | 95.4 |
| 3.65-3.16 | 91.4 | 86.7 | 91.0 |
| 3.16-2.83 | 87.2 | 78.8 | 86.9 |
| 2.83-2.58 | 79.4 | 71.8 | 79.1 |
| 2.58-2.39 | 68.9 | 55.7 | 68.7 |
| 2.39-2.24 | 61.3 | 46.9 | 60.7 |
| 2.24-2.10 | 52.7 | 36.5 | 51.1 |