 |
 |
Hi First thing I would do is throw xia2 at it (from ccp4i2, of course ;-)) and go and have a cup of coffee, and see if the statistics were similar after a Danish pastry. If the stats are better from xia2 than from HKL, go with the xia2 processing, if they are the same (or worse) I'd look more closely at Mark van Raaij's answer (added below) - my _initial_ assumption in this case would be that the crystals aren't isomorphous and the data can't be merged successfully, but it's hard to say much without more information than you've given us. Back to your original processing - What are the stats like for each individual crystal? - Do any of the different datasets merge together with good statistics? - Is the crystal system one where you can have alternative indexing (e.g. hexagonal, tetragonal)? - Pointless should be your friend here Of course, something I would look at would be the various discussions here and elsewhere on the usefulness of Rmerge as an indicator of data quality (hint: it isn't very good compared to other measures). On 11 Apr 2017, at 10:23, Frank von Delft wrote: > Quite a few papers on this in recent years, including from the Sasha Popov (ESRF) and Wayne Hendrickson - don't have references to hand, the Google should sort you out. > > And then of course the CCP4 program BLEND - that's both published and runnable. > > But overall: this is the top problem in data processing at the moment, keeping a lot of smart people very busy. If that makes you feel better ;) Maybe they'll comment. > > > On 11/04/2017 03:46, 高艺娜 wrote: >> Dear all, >> For my crystal, I have to merge 3-4 sets of diffraction data using HKL2000 program to meet requirements for data completeness, but the problem is the Rmerge value was too high to go on every time. >> Did anyone have met the same problem and how to solve it? or some tips for solve this problem？ >> >> All comments will be appreciated! >> >> Best Regards, >> > Hi Gaoyina, > > I guess your crystals ended up in slightly different lattices or with slightly different ordering during flash-cooling. > I this case can see two possible solutions: > - collect more complete data from a single crystal, you may have to "sacrifice" some resolution by collecting shorter images or attenuating the beam somewhat. > - try to flash-cool the crystals in a more reproducible way, so they are more similar to each other. > Another possibility might be that you have too many spot overlaps, in this case you may need to collect thinner-sliced images, offset the detector, put the detector further away (i.e. lower resolution), or measure a crystal with a different orientation (try put the spindle rotation more or less around the long cell axis). > In any of these cases I think you will need to collect more data. > With the current data, in Mosflm data processing you may be able to get more complete data with the SEPARATION CLOSE option, if you did not invoke this already. In other data processing programs there surely are similar options. On the other hand, in the automated data processing pipelines that many synchrotrons use now they are probably already activated when necessary (at least I have rarely been able to improve on automatically processed data when doing it myself later...) > > Greetings, > > Mark J van Raaij > Dpto de Estructura de Macromoleculas > Centro Nacional de Biotecnologia - CSIC > calle Darwin 3 > E-28049 Madrid, Spain > tel. (+34) 91 585 4616 > http://wwwuser.cnb.csic.es/~mjvanraaij >