Hi Frank, On Wed, Jan 28, 2015 at 07:52:39AM +0000, Frank von Delft wrote: > Clemens - what do you mean by "correcting anisostropy"?
Ooops - typo and not quite clear ;-) I mean diffraction anisotropy and correcting for it. SHARP will do that by default if run through autoSHARP and if there is good enough resolution (at least 3.5A). But one can do that in various other ways I guess (including the Diffraction Anisotropy Server at UCLA or various CCP4 programs) - just that I'm most familiar with how SHARP does it. Anyway, the output data (FP/SIGFP as well as map coefficients in eden.mtz) will be corrected for that anisotropy. This means that density-modification will hopefully not get confused by this when working in real-space. I've had cases where this made a significant difference in map quality: convincing me that it is worth while spending time building and improving those initial phases. Cheers Clemens > > On 28/01/2015 07:42, Clemens Vonrhein wrote: > >Dear Nicolas, > > > >On Mon, Jan 26, 2015 at 03:21:00PM +0000, Nicolas Soler wrote: > >>A quick question regarding the density modification interface via > >>the Sharp interface. Which resolution range / radius of the solvent > >>sphere/ ncycles should be used for optimal result? > >I usually don't play around a lot with those parameters - other than > >possibly increasing the number of cycles to something like 100. > > > >>The documentation seems to suggest to restrict yourself up to the > >>resolution where good phasing information is available (6.5A in my > >>case) and I get excellent indicators only if I do that (they become > >>horrible if I use the full resolution range). > >Going from low resolution phase information (say below 4-4.5A) to the > >high resolution limit of your data (even if that is 'only' 3A) is > >notoriously difficult when doing 'only' solvent flattening or > >flipping. To get over that bump, NCS-averaging would be a massive > >help. Or if you have some initial model - maybe a partial MR solution > >or such - you can add this to the mix as well. > > > >The other thing about such low-resolution phase information or data is > >that a lot of the statisitical indicators can become rather noisy and > >much less reliable. > > > >Your statistics will automatically be much worse if you include higher > >resolution data, since you will have many more reflections with poor > >initial phase information going into the same, single value. The > >important thing: do the maps look better than if you use only > >low-resolution data all the way through? > > > >>How about phase extension ? Which parameters would you then use? > >Maybe somebody else has more experience with fine-tuning those > >parameters. For me the some other important things are happening a > >step before that density-modification: > > > >* do the two hands/enantiomoprhs show a significant difference in > > score? > > > >* is the HA model as complete and accurate as possible (using the LLG > > maps in SHARP for checking)? > > > >* is there severe anisotropy in your data? This can be a real pain for > > the density-modification and you should try to correct for it before > > doing any density-modification procedure if possible. > > > >Cheers > > > >Clemens > -- *************************************************************** * Clemens Vonrhein, Ph.D. vonrhein AT GlobalPhasing DOT com * * Global Phasing Ltd. * Sheraton House, Castle Park * Cambridge CB3 0AX, UK *-------------------------------------------------------------- * BUSTER Development Group (http://www.globalphasing.com) ***************************************************************
