Hi Sze Yi As Harry said, in general one should not worry about the refined mosaicity. Especially with suboptimal cryo-conditions crystals can get quite mosaic, so a large refined mosaicity may well be true. Too large a mosaicity will lead to too large integration boxes (noise), but the 3-dimensional spot profiles should take care of that. With too small a mosaicity only the central parts of the spots will be integrated, leading to less accurate results.
The only real problem with a large mosaicity is that the processing program may reject many reflections due to spot overlap, leading to very incomplete data sets. In some cases with extreme mosaicity, I have therefore fixed the mosaicity to a small value, integrating only the central parts of the spots. Although not optimal, this method has produced useful and reasonably complete data sets from diffraction data which otherwise could not be used due to severe incompleteness. Best, Herman Von: CCP4 bulletin board [mailto:[email protected]] Im Auftrag von Harry Powell Gesendet: Dienstag, 23. September 2014 11:11 An: [email protected] Betreff: Re: [ccp4bb] Prediction of Mosaicity Hi Sze Yi The simple answer is that if the indexing solution is correct, you can't, and you shouldn't worry about this. If you are worried, can you send me directly (i.e. *not* to the ccp4bb!) the large date-stamped log file that will be called something like "mosflm_20140820_145228.lp" so that I can have a look at it to see if there's anything obviously wrong with your processing? Provided that the values of refined parameters vary smoothly through the processing run, you shouldn't be too concerned - 0.220mm is quite large, but I see values like this pretty frequently - it's better to consider the normal "Table 1" figures after scaling and merging. Fixing the mosaic spread to a lower value will mean that you are only using parts of the reflections near the middle of their rocking curve to calculate the values (and also not measuring the tails in the third dimension when integrating). The mosaicity estimate is exactly that - it's better than a guess, and in Mosflm is based on analysing the way that the overall intensity measured changes with different values of calculated mosaic spread (in iMosflm, this is done for the first image used in autoindexing). The important thing as far as Mosflm is concerned is that the estimated value should be a good _starting_ value for refining it properly, and is better than just having the value set to zero or some other number that's been plucked out of thin air. Unless you have a crystal with cubic symmetry, there is no reason why the mosaicity should be isotropic (i.e. the same in all directions in the crystal - after all, the cell dimensions will not be the same...), so if you see the value change for different orientations of the crystal, that's what (in general) you should expect. On 23 Sep 2014, at Tue23 Sep 09:38, Lau Sze Yi wrote: Hi, I collected a dataset using our home source (Rigaku/MSC FR-E SuperBright; -AXIS IV++ imaging plate detector) and am processing my data using imosflm. Autoindexing solution is P222, with an estimated mosaicity of 1.20. Mosaicity tends to refine to as high as 1.6, RMS residual ~ 0.220. If I fix the mosaicity at lower values , RMS residual drops. Does this mean that the mosaicity was overestimated in the first place? How do I make sure that mosaicity is not over or under estimated. Appreciate your feedback. Regards, Sze Yi Harry -- Dr Harry Powell, MRC Laboratory of Molecular Biology, Francis Crick Avenue, Cambridge Biomedical Campus, Cambridge, CB2 0QH Chairman of European Crystallographic Association SIG9 (Crystallographic Computing)
