Re: [phenixbb] Similar 2Fo-Fc maps, very different e/A^3 levels (absolute levels)

On 12/03/12 10:38, Sam Stampfer wrote:

Thanks so much for the replies! The structure is being refined at 2.74 A resolution, but I may need to raise the cutoff to 2.9 or so. The crystal was anisotropic.

I would not be surprised to see that the 1 rms level of a 3 A resolution map would be at 0.08 e/A^3. You can't really represent the "peaky" center of atoms with only low resolution Fourier coefficients.

How would I go about properly normalizing the detwinning calculation, if that were the problem?

If your crystal is not twinned then this is not a problem for you. If there is a normalization problem with the detwining code it is up to the authors to figure it out. There is little you can do. However, since I don't expect the variation of a detwinned map to differ much from the untwinned map it appears that you could just multiply your detwinned map by 20 to make it roughly the level of your map #2. Dale Tronrud

Thanks, -Sam

On Mon, Dec 3, 2012 at 1:13 PM, Dale Tronrud mailto:[email protected]> wrote:

I'm not an expert on twining, but I think you should first decide if your crystal is twinned, then use the map that is calculated based on that determination. How the map "looks" is not a very rigorous basis for making a decision about anything, unless you have no bulk solvent regions or your helices "look" left-handed. The fact that your free R does not drop when you refine assuming twining does not encourage me.

The electron density values you quote seem low to me as well (referring to Edward Berry's letter). The expected rms value of a map will depend on resolution. The rms of the map is directly related to the rms of its Fourier coefficients, so when the coefficients get weaker the rms of the map will drop. I'm used to an rms of about 0.36 e/A^3 for a 1.5 A resolution map. For the rms to drop to 0.08 e/A^3 you would have to be at quite a low resolution, but you have not informed us of that.

I suspect that the detwining calculation in Map #1 is not being correctly normalized. If so, the map will appear identical to the correctly normalized map, but only after the contour level has been adjusted. Since this sort of error does not affect most peoples' use of the map this is probably not debugged carefully. The last time I checked, Coot's ncs averaged maps have very odd numbers for their absolute density values. Again, since a spin of the mouse wheel will make the map look as expected no one seems to care enough to figure out the problem.

Dale Tronrud

On 12/03/12 09:09, Sam Stampfer wrote: > Dear Phenix group, > > > > I am trying to decide which electron density map to use for rebuilding > my structure in Coot. Both 2Fo-Fc maps appear quite similar at the 1 > sigma level but the absolute electrons per cubic Angstrom (e/A^3) is > very different. > > > > When contoured at 1.00 rmsd in coot, these are the e/A^3 levels (Coot > calls them absolute levels): > > Map #1 2Fo-Fc: 0.0044 e/A^3 > > Map #2 2Fo-Fc: 0.0832 e/A^3 > > > > Map #1 was generated using a twinning operator that is typically > required for this crystal form. Map #2 was generated without the > twinning operator and it tends to have slightly better 2Fo-Fc density, > and there is a bit more density (or noise?) in the Fo-Fc map. > > > > The structure was refined in Phenix and gave a similar Rfree regardless > of whether a twinning operator was used. > > > > What does it mean to have this 20-fold difference in electrons per cubic > Angstroms for my maps? Which map should I use? > > > > Thanks for your help! > > > > -Sam > > > PS: The Fo-Fc map in Map#2 also contours at a much higher absolute e/A^3 > level. > > > > _______________________________________________ > phenixbb mailing list > [email protected] mailto:[email protected] > http://phenix-onl